What changed in Ai-act-service-desk Frequently Asked…

487 clauses changed: too many to explain in one go, so here is every one of them. Every change was found by comparing the two versions clause by clause, and every quote is checked word for word.

Frequently Asked Questions | AI Act Service Desk

Archived copy · 2026-04-06 Today · 2026-10-15

487 clauses changed: too many to explain in one go, so here is every one of them.

removed
Untitled section
Before
Not in the old version.

After
Removed.
added
- Governance & Enforcement
Before
Not in the old version.

After
- Governance & Enforcement
added
- Governance & Enforcement
Before
Not in the old version.

After
Who is responsible, and who is this aimed at?
added
- Governance & Enforcement
Before
Not in the old version.

After
The AI Office, the European Data Protection Supervisor (EDPS) and the national competent authorities of the Member States are responsible for supervising and enforcing the AI Act.
added
- Governance & Enforcement
Before
Not in the old version.

After
Market surveillance authorities supervise and enforce compliance with the rules for AI systems, including prohibitions and rules for high-risk AI.
added
- Governance & Enforcement
Before
Not in the old version.

After
The EDPS acts as the designated market surveillance authority and notified body specifically responsible for supervising, monitoring, and enforcing compliance for AI systems put into service or used by European Union institutions, bodies, offices, and agencies.
added
- Governance & Enforcement
Before
Not in the old version.

After
The Scientific Panel supports enforcement by supplying independent scientific and technical expertise.
added
- Governance & Enforcement
Before
Not in the old version.

After
Enforcement is aimed at operators subject to the AI Act, including in particular providers and deployers of AI systems, as well as providers of general-purpose AI models.
added
- Governance & Enforcement
Before
Not in the old version.

After
The AI Office has exclusive enforcement powers for certain AI systems and models, namely:
added
- Governance & Enforcement
Before
Not in the old version.

After
- general-purpose AI models, including models with systemic risks;
added
- Governance & Enforcement
Before
Not in the old version.

After
- AI systems based on general-purpose AI models, where the model and the system are developed by the same provider or a provider within the same undertaking as the provider of the underlying GPAI model;
added
- Governance & Enforcement
Before
Not in the old version.

After
- AI systems that constitute, or are integrated into, very large online platforms or very large online search engines designated under Regulation (EU) 2022/2065 (Digital Services Act).
added
- Governance & Enforcement
Before
Not in the old version.

After
- Governance & Enforcement
added
- Governance & Enforcement
Before
Not in the old version.

After
When does enforcement start?
added
- Governance & Enforcement
Before
Not in the old version.

After
Some enforcement powers of the AI Office and the authorities of the Member States will start to apply on 2 August 2026.
added
- Governance & Enforcement
Before
Not in the old version.

After
From that date, the provisions on prohibited AI practices, transparency requirements for certain AI systems, and the rules for general-purpose AI (‘GPAI’) models will be enforceable.
added
- Governance & Enforcement
Before
Not in the old version.

After
However, for certain provisions of the AI Act, enforcement powers will start to apply only when those provisions themselves become applicable.
added
- Governance & Enforcement
Before
Not in the old version.

After
In particular:
added
- Governance & Enforcement
Before
Not in the old version.

After
- Transparency obligations will enter into application and be enforceable from 2 August 2026, although providers of AI systems placed on the market before that date have until 2 December 2026 to comply with marking and detection obligation under Article 50(2).
added
- Governance & Enforcement
Before
Not in the old version.

After
- The new prohibitions relating to the generation or manipulation of non-consensual intimate material and child sexual abuse material will apply from 2 December 2026.
added
- Governance & Enforcement
Before
Not in the old version.

After
- The rules for high-risk AI systems set out in Annex III to the AI Act will apply from 2 December 2027 and the rules for high-risk AI systems embedded into regulated products will apply from 2 August 2028.
added
- General-purpose AI models
Before
Not in the old version.

After
- General-purpose AI models
added
- General-purpose AI models
Before
Not in the old version.

After
How are AI agents addressed within the AI Act?
changed
- General-purpose AI models
Before
From 2 August 2026 onwards, if the AI agent classifies as a high-risk AI system, it is also subject to additional requirements that ensure its safety and trustworthiness for its intended usage (Chapter III AI Act).

After
Furthermore, from 2 Dec 2027 respective 2 Aug 2028 onwards, if the AI agent classifies as a high-risk AI system, it is also subject to additional requirements that ensure its safety and trustworthiness for its intended usage (Chapter III AI Act).
changed
- General-purpose AI models
Before
Furthermore, if the AI agent is intended to interact with natural persons or generate content, transparency rules will apply (Article 50 AI Act) – a Code of Practice to operationalise these rules is in development.

After
From 2 August 2026 onwards, if the AI agent is intended to interact with natural persons or generate content, transparency rules will apply (Article 50 AI Act) – a Code of Practice to operationalise these rules was recently published.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act entered into force on 1 August 2024.

After
Removed.
removed
Untitled section
Before
It follows a staggered entry into application, with some parts already applicable such as certain prohibitions, AI literacy, and rules for general-purpose AI models.

After
Removed.
removed
Untitled section
Before
Other parts of the Act are set to apply on 2 August 2026 and 2 August 2027.

After
Removed.
removed
Untitled section
Before
This progressive roll-out allows us to build on the experience gathered in applying the first part of the rules.

After
Removed.
removed
Untitled section
Before
The Commission is committed to continuously learn and stepping up its efforts.

After
Removed.
removed
Untitled section
Before
This is particularly important in the context of a fast-evolving technology like AI.

After
Removed.
removed
Untitled section
Before
Stakeholder consultations throughout 2025 revealed implementation challenges that need to be addressed so that the AI Act can be successfully rolled-out.

After
Removed.
removed
Untitled section
Before
This proposal puts forwards legislative amendments to that effect and complements ongoing efforts to facilitate compliance with the AI Act, like the launch of an AI Act Service Desk.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
added
- General-purpose AI models
Before
Not in the old version.

After
- General-purpose AI models
added
- General-purpose AI models
Before
Not in the old version.

After
The Commissions’ enforcement powers related to AI Act obligations for providers of the most advanced models enter into application on 2 August 2026.
added
- General-purpose AI models
Before
Not in the old version.

After
What will change then?
added
- General-purpose AI models
Before
Not in the old version.

After
The AI Act foresees a one-year period, from 2 August 2025, in which providers of the most advanced models must comply with their obligations before the Commission’s enforcement powers enter into application on 2 August 2026.
added
- General-purpose AI models
Before
Not in the old version.

After
Providers must, for instance, notify the Commission when they know they are developing one of the most advanced models.
added
- General-purpose AI models
Before
Not in the old version.

After
They must also assess and mitigate systemic risks from those models and most of the relevant providers have signed and are implementing the GPAI Code of Practice to this end.
added
- General-purpose AI models
Before
Not in the old version.

After
Since last year, the Commission’s AI Office has therefore received information and held technical compliance dialogues with providers to improve their practices to assess and mitigate systemic risks.
added
- General-purpose AI models
Before
Not in the old version.

After
The AI Office views these technical compliance dialogues as a first tool of choice to assess compliance and clarify questions to support providers’ efforts.
added
- General-purpose AI models
Before
Not in the old version.

After
So, in one important way, nothing changes in August – the AI Office will continue and, if anything, intensify, the ongoing technical compliance dialogues.
added
- General-purpose AI models
Before
Not in the old version.

After
In preparation for 2 August 2026, the AI Office has also established the structures and competencies needed for using enforcement powers.
added
- General-purpose AI models
Before
Not in the old version.

After
The powers include requesting information, requesting access to a model for evaluations, requiring risk mitigation measures, and issuing fines of up to 3% of global annual turnover or requesting a provider to restrict the making available on the market, withdraw or recall the model.
added
- General-purpose AI models
Before
Not in the old version.

After
The AI Office may use these powers in cases where technical compliance dialogues are not sufficient.
added
- Digital Omnibus
Before
Not in the old version.

After
- Digital Omnibus
added
- Digital Omnibus
Before
Not in the old version.

After
What are the new timelines proposed?
added
- Digital Omnibus
Before
Not in the old version.

After
The proposal acknowledges the challenge that the delay of standards and other support tools cause for the implementation of the AI Act.
added
- Digital Omnibus
Before
Not in the old version.

After
The timeline for the high-risk AI rules is aligned to the availability of standards and other support tools.
added
- Digital Omnibus
Before
Not in the old version.

After
Once the Commission confirms these are sufficiently available, the rules will start to apply after a transition period.
added
- Digital Omnibus
Before
Not in the old version.

After
This flexibility has an end date: the rules for high-risk AI in sensitive areas like employment and law enforcement (Annex III) will in any case apply maximum 16 months later than originally envisaged, the rules for high-risk AI embedded in products like medical devices (Annex I) will apply a maximum 12 months later.
added
- Digital Omnibus
Before
Not in the old version.

After
The proposal also suggests a transition period of 6 months for providers who need to retroactively include technical solutions into their generative AI systems to make them detectable.
added
- Digital Omnibus
Before
Not in the old version.

After
- Digital Omnibus
added
- Digital Omnibus
Before
Not in the old version.

After
How will these changes benefit to businesses?
added
- Digital Omnibus
Before
Not in the old version.

After
According to the Commission’s first estimations, the proposed measures on AI are expected to reduce compliance costs for businesses throughout the EU.
added
- Digital Omnibus
Before
Not in the old version.

After
At the same time, by extending benefits granted to SMEs to include SMCs, the Commission is making implementation easier for an additional 8,250 companies in Europe.
added
- Digital Omnibus
Before
Not in the old version.

After
Overall, the proposals presented by the Commission will help businesses meet their obligations.
added
- Digital Omnibus
Before
Not in the old version.

After
They also open up more opportunities to innovate in the EU, further facilitating the roll-out of the regulatory framework that is designed to create a single market for trustworthy AI.
added
- Digital Omnibus
Before
Not in the old version.

After
- Digital Omnibus
added
- Digital Omnibus
Before
Not in the old version.

After
What are the main changes proposed to the AI Act?
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
According to the Commission’s first estimations, the proposed measures on AI are expected to reduce compliance costs for businesses throughout the EU.

After
Removed.
removed
Untitled section
Before
At the same time, by extending benefits granted to SMEs to include SMCs, the Commission is making implementation easier for an additional 8,250 companies in Europe.

After
Removed.
removed
Untitled section
Before
Overall, the proposals presented by the Commission will help businesses meet their obligations.

After
Removed.
removed
Untitled section
Before
They also open up more opportunities to innovate in the EU, further facilitating the roll-out of the regulatory framework that is designed to create a single market for trustworthy AI.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The proposal acknowledges the challenge that the delay of standards and other support tools cause for the implementation of the AI Act.

After
Removed.
removed
Untitled section
Before
The timeline for the high-risk AI rules is aligned to the availability of standards and other support tools.

After
Removed.
removed
Untitled section
Before
Once the Commission confirms these are sufficiently available, the rules will start to apply after a transition period.

After
Removed.
removed
Untitled section
Before
This flexibility has an end date: the rules for high-risk AI in sensitive areas like employment and law enforcement (Annex III) will in any case apply maximum 16 months later than originally envisaged, the rules for high-risk AI embedded in products like medical devices (Annex I) will apply a maximum 12 months later.

After
Removed.
removed
Untitled section
Before
The proposal also suggests a transition period of 6 months for providers who need to retroactively include technical solutions into their generative AI systems to make them detectable.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The EU AI Act is the world's first comprehensive AI law.

After
Removed.
removed
Untitled section
Before
It aims to promote innovation and uptake of AI, while ensuring a high level of protection of health, safety and fundamental rights, including democracy and the rule of law.

After
Removed.
removed
Untitled section
Before
The uptake of AI systems has a strong potential to bring societal benefits, economic growth and enhance EU innovation and global competitiveness.

After
Removed.
removed
Untitled section
Before
However, in certain cases, the specific characteristics of certain AI systems may create risks related to user safety, including physical safety, and fundamental rights.

After
Removed.
removed
Untitled section
Before
Some powerful AI models that are widely used could also pose systemic risks.

After
Removed.
removed
Untitled section
Before
This leads to legal uncertainty and potentially slower uptake of AI technologies by public authorities, businesses and citizens, due to the lack of trust.

After
Removed.
removed
Untitled section
Before
Disparate regulatory responses by national authorities could risk fragmenting the internal market.

After
Removed.
removed
Untitled section
Before
Responding to these challenges, legislative action was needed to ensure a well-functioning internal market for AI systems and models where both benefits and risks are adequately addressed.

After
Removed.
removed
Related resources
Before
Related resources

After
Removed.
removed
- Artificial Intelligence – Questions and Answers
Before
- Artificial Intelligence – Questions and Answers

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act applies progressively, with a full roll-out by 2 August 2027.

After
Removed.
removed
Untitled section
Before
- The prohibitions, definitions and provisions related to AI literacy became applicable on 2 February 2025;

After
Removed.
removed
Untitled section
Before
- The rules on governance and the obligations for general-purpose AI models became applicable on 2 August 2025;

After
Removed.
removed
Untitled section
Before
- The obligations regarding high-risk systems listed in Annex III, the transparency requirements (Article 50) as well as the measures in support of innovation will apply as of 2 August 2026.

After
Removed.
removed
Untitled section
Before
This is also the date when the enforcement of AI Act will start;

After
Removed.
removed
Untitled section
Before
- The obligations for high-risk AI systems that classify as high-risk because they are embedded in regulated products, listed in Annex I (list of Union harmonisation legislation), will enter into force on 2 August 2027.

After
Removed.
removed
Related resources
Before
Related resources

After
Removed.
removed
- Timeline for the Implementation of the EU AI Act
Before
- Timeline for the Implementation of the EU AI Act

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act is designed as a flexible and future-proof regulation that allows to adapt the rules to the rapid pace of technological development, as well as the potential changes in the use of AI systems and emerging risks.

After
Removed.
removed
Untitled section
Before
While in general the AI Act can only be amended through the legislative procedure, in certain cases, the Commission is empowered to amend certain parts of the AI Act.

After
Removed.
removed
Untitled section
Before
For instance, the following parts of the AI Act can be adapted by the Commission:

After
Removed.
removed
Untitled section
Before
- The list of high-risk use-cases in Annex III.

After
Removed.
removed
Untitled section
Before
The Commission is obliged to carry out a yearly review to assess if changes to the list are needed.

After
Removed.
removed
Untitled section
Before
- The threshold above which general-purpose AI models are presumed to have high impact capabilities and are classified as presenting systemic risks.

After
Removed.
removed
Untitled section
Before
The Commission also regularly assesses if other changes to the AI Act are needed and reports to the European Parliament and the Council.

After
Removed.
removed
Untitled section
Before
Such regular evaluation is foreseen directly in the AI Act.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act does not apply to all AI solutions, but only to those that fulfil the definition of an ‘AI system’ within the meaning of Article 3(1) AI Act.

After
Removed.
removed
Untitled section
Before
The AI Act follows a risk-based approach and introduces rules for AI systems based on the level of risk they can pose.

After
Removed.
removed
Untitled section
Before
Any AI practices with an unacceptable risk to health, safety or fundamental rights enshrined in the Charter of Fundamental Rights are prohibited (e.g.

After
Removed.
removed
Untitled section
Before
AI systems used to detect emotions of employees at work, except medical and safety reasons; certain social scoring practices).

After
Removed.
removed
Untitled section
Before
AI systems with high risk for health, safety or fundamental rights need to meet certain requirements to make sure they are safe and trustworthy (e.g.

After
Removed.
removed
Untitled section
Before
AI systems used at border control management; law enforcement, or autonomous vehicles could be examples of high-risk AI systems).

After
Removed.
removed
Untitled section
Before
Certain AI systems need to meet transparency requirements (e.g., deep fakes will have to be labelled as AI-generated; chatbots should inform that a person is not communicating with a human).

After
Removed.
removed
Untitled section
Before
All other AI systems remain unregulated and can be placed on the market, put into service or used in the EU without any requirements – at the time of preparation of the proposal of AI Act, it was estimated that these would be 85%.

After
Removed.
removed
Related resources
Before
Related resources

After
Removed.
removed
- Artificial Intelligence – Questions and Answers
Before
- Artificial Intelligence – Questions and Answers

After
Removed.
removed
- Artificial Intelligence – Questions and Answers
Before
- The Commission publishes guidelines on AI system definition to facilitate the f…

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act does not automatically apply to all AI systems placed on the market before its application date.

After
Removed.
removed
Untitled section
Before
Instead, compliance obligations are phased in depending on the category and whether the system undergoes significant modifications.

After
Removed.
removed
Untitled section
Before
The AI Act will apply from 31 December 2030 to AI systems which are components of the large-scale IT systems established by the legal acts listed in Annex X that have been placed on the market or put into service before 2 August 2027.

After
Removed.
removed
Untitled section
Before
In case such large-scale IT systems are evaluated or the legal acts in Annex X are replaced or amended before 31 December 2030, the AI Act shall be taken into account.

After
Removed.
removed
Untitled section
Before
The AI Act will also apply to high-risk AI systems that have been placed on the market or put into service before 2 August 2026 only in case those systems are significantly modified.

After
Removed.
removed
Untitled section
Before
The AI Act will also apply from 31 December 2030 to high-risk AI systems that have been placed on the market or put into service before 2 August 2026 and are (intended to be) used by public authorities.

After
Removed.
removed
Untitled section
Before
Lastly, the AI Act will apply from 2 August 2027 to general-purpose AI models that have been placed on the market before 2 August 2025.

After
Removed.
removed
Untitled section
Before
However, the rules on prohibited AI practices (Article 5) apply to all AI systems, without taking into account the date of their placement on the market.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Focusing specifically on the building phase of the AI system, Recital 12 of the AI Act further clarifies that ‘[t]he techniques that enable inference while building an AI system include machine learning approaches that learn from data how to achieve certain objectives, and logic- and knowledge-based approaches that infer from encoded knowledge or symbolic representation of the task to be solved.’ Those techniques should be understood as ‘AI techniques’.

After
Removed.
removed
Untitled section
Before
In addition to various machine learning approaches, that are often understood as an AI technique, the second category of AI techniques mentioned in Recital 12 of the AI Act are ‘logic- and knowledge-based approaches that infer from encoded knowledge or symbolic representation of the task to be solved’.

After
Removed.
removed
Untitled section
Before
Instead of learning from data, these AI systems learn from knowledge including rules, facts and relationships encoded by human experts.

After
Removed.
removed
Untitled section
Before
Based on the knowledge encoded by human experts, these systems can ‘reason’ via deductive or inductive engines or using operations such as sorting, searching, matching and chaining.

After
Removed.
removed
Untitled section
Before
By using logical inference to draw conclusions, such systems apply formal logic, predefined rules or ontologies to new situations.

After
Removed.
removed
Untitled section
Before
Logic and knowledge based approaches include, for instance, knowledge representation, inductive (logic) programming, knowledge bases, inference and deductive engines, (symbolic) reasoning, expert systems and search and optimisation methods.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act distinguishes between an AI system, defined in Article 3(1) and a general-purpose AI model defined in Article 3(63).

After
Removed.
removed
Untitled section
Before
The AI Act refers to AI models as those that ‘are essential components of AI systems.

After
Removed.
removed
Untitled section
Before
They do not constitute AI systems on their own.

After
Removed.
removed
Untitled section
Before
AI models require the addition of further components, such as for example a user interface, to become AI systems.

After
Removed.
removed
Untitled section
Before
AI models are typically integrated into and form part of AI systems’ (Recital 97). ‘Large generative AI models are a typical example for a general-purpose AI model, given that they allow for flexible generation of content, such as in the form of text, audio, images or video, that can readily accommodate a wide range of distinctive tasks’ (Recital 99).

After
Removed.
removed
Untitled section
Before
The AI Act defines an AI system as a ‘machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments’ (Article 3(1), Recital 12).

After
Removed.
removed
Untitled section
Before
For more information, please refer to the Guidelines on the definition of an artificial intelligence system.

After
Removed.
removed
Related resources
Before
Related resources

After
Removed.
removed
- Artificial Intelligence – Questions and Answers
Before
- Artificial Intelligence – Questions and Answers

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act follows a risk-based approach, classifying AI systems into four different risk categories: unacceptable risk, high risk, transparency risk, and minimal to no risk.

After
Removed.
removed
Untitled section
Before
For the unacceptable risk category, the AI Act lists specific practices that are prohibited (Article 5 of the AI Act).

After
Removed.
removed
Untitled section
Before
The high-risk AI systems are defined in accordance with Article 6 of the AI Act in conjunction with Annex I (list of Union harmonisation legislation) and Annex III of the AI Act.

After
Removed.
removed
Untitled section
Before
Annex III comprises eight areas in which the use of AI can be particularly sensitive and lists concrete use cases for each area that are assessed by the co-legislator as posing significant risks to health, safety and fundamental rights.

After
Removed.
removed
Untitled section
Before
For certain AI systems where transparency is particularly important, rules on transparency are foreseen.

After
Removed.
removed
Untitled section
Before
All the other AI systems are considered as “minimal to no risk” AI systems and the AI act does not foresee any obligations for them.

After
Removed.
removed
Untitled section
Before
Voluntarily, providers of those systems may choose to apply the requirements for trustworthy AI and adhere to voluntary codes of conduct.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Depending on the level of risk, obligations and requirements related to specific AI systems will vary.

After
Removed.
removed
Untitled section
Before
Providers of high-risk AI systems need to ensure that their high-risk AI system complies with the requirements of the AI Act before the system is placed on the market or put into service.

After
Removed.
removed
Untitled section
Before
Examples of such requirements include having a risk management system in place, recording logs, and ensuring qualitative data and data governance, as well as human oversight.

After
Removed.
removed
Untitled section
Before
In addition, providers will need to register their system in the EU database and accompany the system with instructions for its use for deployers.

After
Removed.
removed
Untitled section
Before
Providers of transparency-risk systems, such as chatbots, AI social companions or deep fakes, need to comply with certain transparency obligations.

After
Removed.
removed
Untitled section
Before
This includes, for example, informing natural persons that they are interacting with an AI system or ensuring that AI generated content can be detected as such.

After
Removed.
removed
Untitled section
Before
Deployers of systems that can generate or manipulate content need to disclose that the content is generated by AI.

After
Removed.
removed
Untitled section
Before
The AI Act does not prescribe obligations for minimal-risk AI systems, but Member States can facilitate the drawing up of voluntary codes of conduct for AI systems that are not high-risk.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
In principle, the evolution of an AI system does not have legal implications.

After
Removed.
removed
Untitled section
Before
However, in the case of high-risk AI systems, such developments should be anticipated and addressed within the risk management framework and related compliance obligations.

After
Removed.
removed
Untitled section
Before
Where an AI system undergoes substantial modifications, this may give rise to legal consequences, including the need for a renewed conformity assessment under the AI Act.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act does not require any particular internal governance within the company.

After
Removed.
removed
Untitled section
Before
However in line with Article 17(1)(m) the providers of high-risk AI systems should put a quality management system in place, that should include an accountability framework, setting up the responsibilities of the management and other staff with regard to all the aspects related to the quality management system listed in Article 17 of the AI Act.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The EU Funding & Tenders Portal of the European Commission is the central place to find any funding opportunities from the EU, including AI-relevant ones.

After
Removed.
removed
Untitled section
Before
Of particular interest among the funding programmes are Horizon Europe and Digital Europe Programme.

After
Removed.
removed
Untitled section
Before
Within Horizon Europe, there are different areas of interest, namely Cluster 4 and the European Innovation Council.

After
Removed.
removed
Untitled section
Before
Other areas such as the bottom-up European Research Council may be also relevant.

After
Removed.
removed
Untitled section
Before
The GenAI4EU flagship initiative offers ample opportunities to develop and deploy trustworthy, generative AI in Europe’s strategic sectors.

After
Removed.
removed
Untitled section
Before
With nearly 700 million euros committed, there are plenty of opportunities to help Europe become more competitive and innovative.

After
Removed.
removed
Untitled section
Before
A dedicated website gives an overview of the different call for proposals available under Horizon Europe and the Digital Europe Programme.

After
Removed.
removed
Related resources
Before
Related resources

After
Removed.
removed
- EU Funding & Tenders Portal
Before
- EU Funding & Tenders Portal

After
Removed.
removed
- GenAI4EU: Funding opportunities to boost Generative AI “made in Europe”
Before
- GenAI4EU: Funding opportunities to boost Generative AI “made in Europe”

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The concept of AI literacy mentioned in Article 4 of the AI Act relies on the definition of the term given in Article 3(56) of the AI Act, according to which: ‘AI literacy’ means skills, knowledge and understanding that allow providers, deployers and affected persons, taking into account their respective rights and obligations in the context of this Regulation, to make an informed deployment of AI systems, as well as to gain awareness about the opportunities and risks of AI and possible harm it can cause.

After
Removed.
removed
Untitled section
Before
Read the complete AI Literacy FAQ here.

After
Removed.
removed
Related resources
Before
Related resources

After
Removed.
removed
- AI Literacy - Questions & Answers
Before
- AI Literacy - Questions & Answers

After
Removed.
removed
- Living repository to foster learning and exchange on AI literacy
Before
- Living repository to foster learning and exchange on AI literacy

After
Removed.
removed
- Third AI Pact webinar on AI literacy
Before
- Third AI Pact webinar on AI literacy

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
There is no single approach that works for all.

After
Removed.
removed
Untitled section
Before
It depends on the type of AI system that the company deals with and the knowledge of the staff in question.

After
Removed.
removed
Untitled section
Before
There is, however, no obligation for external training or external certification.

After
Removed.
removed
Untitled section
Before
Please see further AI Literacy - Questions & Answers.

After
Removed.
removed
Related resources
Before
Related resources

After
Removed.
removed
- AI Literacy - Questions & Answers
Before
- AI Literacy - Questions & Answers

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act does not require that companies designate AI Officers.

After
Removed.
removed
Untitled section
Before
However, the AI Act requires that providers and deployers of AI systems ensure, to their best extent, a sufficient level of AI literacy of their staff and other people dealing with the operation and use of AI systems on their behalf.

After
Removed.
removed
Untitled section
Before
The Commission published a living repository to foster learning and exchange on AI literacy, Questions and Answers on AI literacy, and a recorded webinar on AI literacy.

After
Removed.
removed
Untitled section
Before
The repository will be expanded.

After
Removed.
removed
Untitled section
Before
There are many cost-free online training courses.

After
Removed.
removed
Untitled section
Before
The choice depends on the individual skills of the staff in question.

After
Removed.
removed
Related resources
Before
Related resources

After
Removed.
removed
- Living repository to foster learning and exchange on AI literacy
Before
- Living repository to foster learning and exchange on AI literacy

After
Removed.
removed
- AI Literacy - Questions & Answers
Before
- AI Literacy - Questions & Answers

After
Removed.
removed
- Third AI Pact webinar on AI literacy
Before
- Third AI Pact webinar on AI literacy

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The prohibitions of Article 5 AI Act include certain manipulative practices, social scoring, predictive policing based solely on profiling, scraping the internet and CCTV material to build-up or expand facial recognition databases, emotion recognition at education and workplace, biometric categorisation to infer sensitive attributes such as political opinion or sexual orientation and real-time remote biometric identification for law enforcement purposes at publicly accessible spaces (with some limited exceptions).

After
Removed.
removed
Untitled section
Before
For more information, please refer to the Guidelines on prohibited artificial intelligence practices.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The prohibitions of Article 5 include certain manipulative practices, social scoring, predictive policing based solely on profiling, scraping the internet and CCTV material to build-up or expand facial recognition databases, emotion recognition at education and workplace, biometric categorisation to infer sensitive attributes such as political opinion or sexual orientation and real-time remote biometric identification for law enforcement purposes at publicly accessible spaces (with some limited exceptions).

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Guidelines were issued in February 2025.

After
Removed.
removed
Related resources
Before
Related resources

After
Removed.
removed
Related resources
Before
- Commission publishes the Guidelines on prohibited artificial intelligence (AI) …

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The Commission is currently preparing guidelines on the classification of high-risk AI systems, which should be published in February 2026.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The placing on the market, the putting into service for this specific purpose, or the use of AI systems to infer emotions of a natural person in the areas of workplace and education institutions, is prohibited.

After
Removed.
removed
Untitled section
Before
However, this prohibition does not apply to the use of an AI system when the AI system is intended to be put in place or into the market for medical or safety reasons.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Annex III comprises eight areas and lists specific use-cases for each, which the EU co-legislator has assessed as posing a high risk of harm to health and safety or to fundamental rights.

After
Removed.
removed
Untitled section
Before
In simple terms, the areas are enumerated as points in Annex III (e.g. point 1 – Biometrics, point 2 – Critical infrastructure, point 3 – Education and vocational training, etc.).

After
Removed.
removed
Untitled section
Before
The use-cases are listed under each area.

After
Removed.
removed
Untitled section
Before
For example, point 6 of Annex III covers the area of law enforcement, which includes five use-cases ((a)–(e)).

After
Removed.
removed
Untitled section
Before
To be classified as high-risk, an AI system must fall under one of the areas of Annex III and correspond to one of the use-cases listed therein.

After
Removed.
removed
Untitled section
Before
In other words, not all AI systems used in a particular area (such as law enforcement) are automatically considered high-risk; only those falling within the specific use-cases identified in Annex III qualify.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Article 16 of the AI Act gives an overview of the obligations that providers of high-risk AI systems need to follow in order to comply with the AI Act.

After
Removed.
removed
Untitled section
Before
Prior to placing their high-risk AI system or putting it into services, providers of such systems need to ensure that their high-risk AI system is compliant with Articles 8-15 of the AI Act and has undergone a conformity assessment.

After
Removed.
removed
Untitled section
Before
This conformity should be demonstrated upon a reasonable request by a competent national authority.

After
Removed.
removed
Untitled section
Before
Following this conformity assessment, providers need to draw up an EU declaration of conformity and affix the CE marking to the system (or on its packaging or accompanying documentation), as well as indicate their contact information on the high-risk AI system so they can be contacted.

After
Removed.
removed
Untitled section
Before
Providers also need to register their high-risk AI system in the EU database.

After
Removed.
removed
Untitled section
Before
In addition, providers need to comply with Articles 17-20 AI Act by setting up a quality management system, keep documentation for 10 years, keep automatically generated logs by the high-risk AI system, and take necessary corrective actions and provide information related to these actions.

After
Removed.
removed
Untitled section
Before
Finally, providers of high-risk AI systems need to ensure that their system complies with relevant accessibility requirements.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Location data is normally not biometric.

After
Removed.
removed
Untitled section
Before
Accordingly, it does not fall within the high-risk use cases listed in point 1 of Annex III, nor within the prohibited practices set out in Article 5(1)(g).

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Most biometric AI systems are not prohibited under the AI Act.

After
Removed.
removed
Untitled section
Before
The prohibited AI practices regarding biometric systems are limited and include in particular: emotion recognition in the workplace, certain biometric categorisation systems, and real-time remote biometric identification in publicly accessible spaces for law enforcement purposes.

After
Removed.
removed
Untitled section
Before
These are subject to narrowly defined exceptions.

After
Removed.
removed
Untitled section
Before
There are, however, some additional rules that apply to permitted biometric systems.

After
Removed.
removed
Untitled section
Before
In particular, certain biometric applications are considered high-risk (remote biometric identification systems, certain AI systems intended for biometric categorisation, and AI systems intended for emotion recognition).

After
Removed.
removed
Untitled section
Before
The AI Act sets out specific requirements for high-risk AI systems.

After
Removed.
removed
Untitled section
Before
These requirements relate to, among others, data and data governance, documentation and record-keeping, transparency and the provision of information to users, human oversight, robustness, accuracy, and security.

After
Removed.
removed
Untitled section
Before
The use of AI systems for biometric verification - that is, confirming that an individual is who they claim to be - is not prohibited under the AI Act and does not fall within the category of high-risk systems.

After
Removed.
removed
Untitled section
Before
For more information, please refer to Guidelines on prohibited artificial intelligence practices.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The EU AI Act emphasises the importance of human oversight for high-risk AI systems.

After
Removed.
removed
Untitled section
Before
This means that these systems must be designed and developed so that humans can monitor them while they are in use.

After
Removed.
removed
Untitled section
Before
Appropriate tools and measures should be in place to help people supervise AI effectively, ensuring that any risks to health, safety, or fundamental rights are prevented or minimised.

After
Removed.
removed
Untitled section
Before
When the system is in use, human oversight also includes appropriate measures to be implemented by the deployer.

After
Removed.
removed
Untitled section
Before
Those measures should be described in the instructions of use and be effective and commensurate with the risk, the level of autonomy of the system and the context of use of the high-risk AI system.

After
Removed.
removed
Untitled section
Before
Deployers should implement human oversight measures and assign this task to natural persons who have the necessary competence, training and authority, as well as the necessary support.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act introduces a number of support measures tailored to small and medium-sized enterprises (SMEs).

After
Removed.
removed
Untitled section
Before
These include the possibility to prepare simplified technical documentation and to implement simplified quality management systems for high-risk AI systems.

After
Removed.
removed
Untitled section
Before
SMEs are also granted free-of-charge access to AI regulatory sandboxes, enabling them to test and develop AI systems in a controlled environment.

After
Removed.
removed
Untitled section
Before
In addition, the European Commission and national authorities are required under Article 62 to provide specific support measures to SMEs.

After
Removed.
removed
Untitled section
Before
Their interests are further represented through a special membership category in the advisory forum, ensuring that their needs are taken into account in the regulatory process.

After
Removed.
removed
Untitled section
Before
Finally, SMEs benefit from special consideration when penalties are imposed, reflecting their particular position in the market and the potential impact of sanctions.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
When a general-purpose AI model is integrated into or forms part of an AI system, the resulting system should be considered a general-purpose AI system if it gains the capability to serve multiple purposes through this integration.

After
Removed.
removed
Untitled section
Before
General-purpose AI systems can be used directly or embedded in other AI systems.

After
Removed.
removed
Untitled section
Before
Given their versatility, general-purpose AI systems may be used as high-risk AI systems in their own right or as components of other high-risk AI systems.

After
Removed.
removed
Untitled section
Before
To ensure a fair distribution of responsibilities throughout the AI value chain, providers of general-purpose AI systems should cooperate closely with providers of relevant high-risk AI systems, unless otherwise specified in this Regulation.

After
Removed.
removed
Untitled section
Before
This cooperation should enable high-risk AI system providers to comply with their obligations under this Regulation and facilitate interactions with the competent authorities established by the Regulation.

After
Removed.
removed
Related resources
Before
Related resources

After
Removed.
removed
- Questions and Answers on the Code of Practice for General-Purpose AI
Before
- Questions and Answers on the Code of Practice for General-Purpose AI

After
Removed.
removed
- General-Purpose AI Models in the AI Act – Questions & Answers
Before
- General-Purpose AI Models in the AI Act – Questions & Answers

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
While AI agents are not a separate category under the AI Act, they may have to comply with the requirements for AI systems and/or the obligations for providers of general-purpose AI models.

After
Removed.
removed
Untitled section
Before
Particularly relevant are the AI Act’s prohibitions of harmful manipulation or exploitation of vulnerabilities, which can require technical safeguards in the design of an agentic AI system.

After
Removed.
removed
Untitled section
Before
If the agentic AI system classifies as high-risk, it is subject to additional requirements that ensure its safety and trustworthiness.

After
Removed.
removed
Untitled section
Before
Furthermore, if the agentic AI system is intended to interact with natural persons or generate content, transparency rules apply (for further information, see these FAQs and this consultation).

After
Removed.
removed
Untitled section
Before
As regards general-purpose AI models, factors like the level of autonomy or tool use of a model can be decisive in its designation as having systemic risk.

After
Removed.
removed
Untitled section
Before
Providers of general-purpose AI models with systemic risk are subject to risk management obligations, which include considerations regarding the model’s autonomous capabilities and its agentic use.

After
Removed.
removed
Untitled section
Before
Given that the developments are recent and fast evolving, the European Commission’s regulatory considerations are only preliminary at this stage, also in light of the fact that the understanding of what constitutes an AI agent varies greatly.

After
Removed.
removed
Untitled section
Before
The European Commission continues to closely monitor the development of AI agents and, if need be, will consider actions in that regard.

After
Removed.
removed
Untitled section
Before
For example, the AI Office’s recent call for tenders on technical assistance for AI safety includes a dedicated lot to evaluating safety and security of AI agents.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act regulates amongst others general-purpose AI models that are placed on the Union market, including their development and use.

After
Removed.
removed
Untitled section
Before
For further information under which conditions an AI model is considered ‘general-purpose’, see the question ‘When does a model qualify as a general-purpose AI model?’ and Section 2.1 of the Guidelines on the scope of the obligations for general-purpose AI models.

After
Removed.
removed
Untitled section
Before
‘Placing on the market’ means the first supply of an AI system or a general-purpose AI model for distribution or use on the Union market during a commercial activity, whether in return for payment or free of charge (Article 3(9)(10) AI Act).

After
Removed.
removed
Untitled section
Before
Some examples of what constitutes a placing on the market can be found in Section 3.1.2 of the Guidelines on the scope of the obligations for general-purpose AI models.

After
Removed.
removed
Untitled section
Before
An open-source release can constitute a placing on the market.

After
Removed.
removed
Untitled section
Before
Notably, the commercial supply of the model can be free of charge (Article 3(10) AI Act).

After
Removed.
removed
Untitled section
Before
However, providers of an open-source general-purpose AI model that does not present systemic risk are exempt from certain obligations.

After
Removed.
removed
Untitled section
Before
For further information, see the question ‘How does the AI Act apply to general-purpose AI models released as open-source?’ and Section 4 of the Guidelines on the scope of the obligations for general-purpose AI models.

After
Removed.
removed
Untitled section
Before
The internal use of a general-purpose AI model constitutes a placing on the market at least if it is essential for providing a product or service to third parties on the Union market or if it affects the rights of natural persons in the Union.

After
Removed.
removed
Untitled section
Before
Further examples are provided in Section 3.1.2 of the Guidelines on the scope of the obligations for general-purpose AI models.

After
Removed.
removed
Untitled section
Before
AI models that are specifically developed and put into service for the sole purpose of scientific research and development are out of scope of the AI Act (Article 2(6) AI Act).

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Article 3(63) AI Act defines a ‘general-purpose AI model’ as ‘an AI model, including where such an AI model is trained with a large amount of data using self-supervision at scale, that displays significant generality and is capable of competently performing a wide range of distinct tasks regardless of the way the model is placed on the market and that can be integrated into a variety of downstream systems or applications, except AI models that are used for research, development or prototyping activities before they are placed on the market’.

After
Removed.
removed
Untitled section
Before
This definition lists in a general manner factors that determine whether a model is a general-purpose AI model.

After
Removed.
removed
Untitled section
Before
Nevertheless, it does not set out specific criteria that potential providers can use to assess whether their model is a general-purpose AI model.

After
Removed.
removed
Untitled section
Before
Section 2.1 of the Guidelines on the scope of obligations for general-purpose AI models provide an indicative criterion: a model is considered a general-purpose AI model if the computational resources used for its training exceeds 10^23 floating point operations and it can generate language (text or audio), text-to-image, or text-to-video.

After
Removed.
removed
Untitled section
Before
This compute threshold corresponds to the typical amount of compute used to train models with one billion parameters on large datasets.

After
Removed.
removed
Untitled section
Before
Moreover, the chosen modalities enable flexible content generation capable of accommodating a wide range of distinct tasks.

After
Removed.
removed
Untitled section
Before
However, this is not an absolute rule — models meeting this criterion may exceptionally not qualify as general-purpose AI models if they lack significant generality, while models below this threshold may still be general-purpose AI models if they display significant generality and can competently perform a wide range of tasks.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
A general-purpose AI model is classified as a general-purpose AI model with systemic risk if it meets either of the following two conditions.

After
Removed.
removed
Untitled section
Before
First, if it has capabilities that match or exceed those of the most advanced models.

After
Removed.
removed
Untitled section
Before
The AI Act presumes models trained with a cumulative amount of computational resources that exceeds 10^25 floating point operations to have such capabilities (Article 51(2) AI Act).

After
Removed.
removed
Untitled section
Before
The threshold is set on the basis that the use of this amount of compute indicates that the model could significantly impact the Union market due to its reach or potential negative effects on public health, safety, security, fundamental rights, or society.

After
Removed.
removed
Untitled section
Before
The European Commission shall adjust this threshold when necessary to account for technological developments.

After
Removed.
removed
Untitled section
Before
Second, the European Commission can designate a model as a general-purpose AI model with systemic risk, either through its own initiative or following a qualified alert from the scientific panel, if the model has capabilities or an impact equivalent to those of the most advanced models (Article 51(1)(b) AI Act).

After
Removed.
removed
Untitled section
Before
This accounts for the fact that there are models that may present systemic risks despite not meeting the compute threshold.

After
Removed.
removed
Untitled section
Before
For a model meeting either condition, its provider must comply with additional obligations, including assessing and mitigating systemic risks (see the question ‘What are the additional requirements for general-purpose AI models with systemic risk?’).

After
Removed.
removed
Untitled section
Before
Providers of models trained with less than 10^25 floating point operations of cumulative compute are not expected to notify the European Commission of their model and comply with the safety and security obligations, unless their model was designated as presenting systemic risks by the European Commission.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Training a general-purpose AI model (with or without systemic risk) requires considerable upfront allocation of compute resources.

After
Removed.
removed
Untitled section
Before
Therefore, providers are able to know whether their model will meet the indicative criterion for constituting a general-purpose AI model, given in Section 2.1 of the Guidelines on the scope of the obligations for general-purpose AI models, or the cumulative training compute threshold in Article 51(2) of the AI Act, before the training is complete (recital 112 AI Act).

After
Removed.
removed
Untitled section
Before
In particular, providers are expected to estimate whether their model will be a general-purpose AI model (trained with more than 10^23 floating point operations of cumulative compute) and a general-purpose AI model with systemic risk (trained with more than 10^25 floating point operations of cumulative compute) before the start of a large pre-training run.

After
Removed.
removed
Untitled section
Before
Once a model exceeds the cumulative training compute threshold of 10^25 floating point operations, set in Article 51(2) of the AI Act, or the provider knows that the model will exceed this threshold, the provider has two weeks to notify the European Commission (Article 52 AI Act).

After
Removed.
removed
Untitled section
Before
For further information on the notification see the question ‘If a general-purpose AI model is above the compute threshold for such models with systemic risk, is it always considered to have systemic risk?’.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Providers may choose their method to calculate/estimate the cumulative training compute in floating point operations.

After
Removed.
removed
Untitled section
Before
The Guidelines on the scope of the obligations for general-purpose AI models, in their Annex, offer two possible approaches: architecture-based and hardware-based; as well as general principles for what should be included/excluded in the calculation.

After
Removed.
removed
Untitled section
Before
The compute used to generate synthetic data — artificially created training data produced by algorithms (including other models) rather than collected from real-world sources — must be included in the cumulative training compute calculation, additionally to the compute used to train on the synthetic data, if the synthetic data is not publicly accessible.

After
Removed.
removed
Untitled section
Before
However, the cumulative training compute of the parent model (i.e., the model generating the synthetic data) does not need to be included in the calculation/estimation.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Providers may contest the automatic classification as a general-purpose AI model with systemic risk, when meeting the compute threshold in Article 51(2) of the AI Act.

After
Removed.
removed
Untitled section
Before
For instance, they may rebut the presumption that their model has high-impact capabilities by providing evidence that their model’s capabilities do not match or exceed those recorded in the most advanced models.

After
Removed.
removed
Untitled section
Before
Section 2.3 of the Guidelines on the scope of the obligations for general-purpose AI models provides further guidance on how to do this.

After
Removed.
removed
Untitled section
Before
If the European Commission accepts the provider’s arguments, the model will cease to be classified as systemic risk, thereby reducing the obligations for the provider with regard to this specific model from the moment of acceptance onwards.

After
Removed.
removed
Untitled section
Before
In line with the collaborative and proportionate application of the AI Act, and the provider’s right to good administration, the European Commission may take into account new information submitted by the provider after the notification, once it becomes available.

After
Removed.
removed
Untitled section
Before
Furthermore, the European Commission’s decision to accept or reject the provider’s arguments may take into account uncertainties in the information provided and outline conditions for potential future reassessment accordingly.

After
Removed.
removed
Untitled section
Before
However, the obligations for providers of general-purpose AI models with systemic risk apply from the moment the model meets the condition of Article 51(1)(a) of the AI Act (i.e., high-impact capabilities, presumed by meeting the compute threshold).

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Generally, no.

After
Removed.
removed
Untitled section
Before
Safety measures may be relied upon to mitigate systemic risks stemming from the provider’s general-purpose AI model with systemic risk.

After
Removed.
removed
Untitled section
Before
However, mitigations do not change the fact that the model presents (acceptable) systemic risk.

After
Removed.
removed
Untitled section
Before
If the model does not present systemic risks, even without implemented safety and security measures, providers may contest the automatic classification pursuant to Article 52(2) of the AI Act (see the question ‘If a general-purpose AI model is above the compute threshold for such models with systemic risk, is it always considered to have systemic risk?’).

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The European Commission considers any subsequent development of the model downstream of its large pre-training run to be part of the same model lifecycle and therefore part of the same model, if performed by the same provider or on behalf of the same provider (see Section 2.2 of the Guidelines on the scope of the obligations for general-purpose AI models).

After
Removed.
removed
Untitled section
Before
Different considerations apply if another actor modifies the model (see the question ‘If a general-purpose AI model is modified by another actor than the original provider, does the modifying entity become the provider?’ and Section 3.2 of the Guidelines on the scope of the obligations for general-purpose AI models).

After
Removed.
removed
Untitled section
Before
This understanding of a model was chosen by the European Commission after careful consideration.

After
Removed.
removed
Untitled section
Before
In modern AI development, it is common to make small iterative changes to AI models after market placement.

After
Removed.
removed
Untitled section
Before
Providers also often perform more significant post-training to their model already placed on the market and the resulting altered model gets marketed under a new name, colloquially being referred to as a ‘new model’.

After
Removed.
removed
Untitled section
Before
As our understanding of the technology is still evolving, currently it is hard to define a good boundary in what counts as a relevant or irrelevant modification for the purpose of delineating subsequent model versions into distinct models.

After
Removed.
removed
Untitled section
Before
Therefore, for now, the European Commission considers every model version descended from the same large pre-training run by the same provider to be the same model to provide legal certainty.

After
Removed.
removed
Untitled section
Before
As a result, not every ‘release’ that is branded as such by providers is necessarily a new model for the purposes of the AI Act, following the interpretation in the guidelines.

After
Removed.
removed
Untitled section
Before
Non-public technical details about the training process may be needed to determine whether a ‘release’ constitutes a new model.

After
Removed.
removed
Untitled section
Before
As models placed on the market before 2 August 2025 only need to be compliant with the AI Act starting from 2 August 2027 (Article 111(3) AI Act), it is possible that this transitional period still applies to a ‘release’ happening now.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Different to the provisions on AI systems, which cover multiple different roles such as the provider, deployer, importer, or distributor, the provisions on general-purpose AI models only regulate the ‘provider’.

After
Removed.
removed
Untitled section
Before
A ‘provider’ in this context is whoever develops a general-purpose AI model or that has a general-purpose AI model developed, and places it on the market under its own name or trademark, whether for payment or free of charge (Article 3(3) AI Act).

After
Removed.
removed
Untitled section
Before
Importantly, ‘places it on the market’ refers to the first commercial supply of the model for distribution or use on the Union market (Article 3(9)(10) AI Act).

After
Removed.
removed
Untitled section
Before
For example, an employer that provides its employees with licenses for a model already available on the Union market would neither be considered the provider of that model nor be required to verify its compliance under the AI Act.

After
Removed.
removed
Untitled section
Before
However, if a downstream actor integrates a model that was already made available on the Union market into its AI system, it may face obligations under the AI Act’s provisions on AI systems.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The European Commission considers a downstream modifier to become the provider of the modified general-purpose AI model only if the modification leads to a significant change in the model’s generality, capabilities, or systemic risk.

After
Removed.
removed
Untitled section
Before
An indicative criterion for this is that the training compute used for the modification is greater than a third of the training compute of the original model (see the Annex of the Guidelines on the scope of the obligations for general-purpose AI models for how ‘training compute’ should be understood).

After
Removed.
removed
Untitled section
Before
However, if the downstream modifier cannot be expected to know this value and cannot estimate it then the threshold should be replaced as follows.

After
Removed.
removed
Untitled section
Before
If the original model is a general-purpose AI model with systemic risk, the threshold should be replaced with a third of the threshold for a model being presumed to have high-impact capabilities (i.e., currently 10^25 floating point operations).

After
Removed.
removed
Untitled section
Before
Otherwise, it should be replaced with a third of the threshold for a model being presumed to be a general-purpose AI model (i.e., currently 10^23 floating point operations, see Section 2.1 of the Guidelines on the scope of the obligations for general-purpose AI models).

After
Removed.
removed
Untitled section
Before
This threshold is based on the expectation that a model that is modified with this amount of compute will display a significant change which warrants the downstream modifier being subject to the obligations for providers or general-purpose AI models and potentially of general-purpose AI models with systemic risk.

After
Removed.
removed
Untitled section
Before
While currently few modifications may meet this criterion, the number of downstream modifiers that become providers of general-purpose AI models may increase over time as the compute used to modify models increases.

After
Removed.
removed
Untitled section
Before
The criterion is thus primarily forward-looking, and in line with the risk-based approach of the AI Act.

After
Removed.
removed
Untitled section
Before
Therefore, the European Commission’s approach may change in the future as technology and the market evolve.

After
Removed.
removed
Untitled section
Before
For further information see Section 3.2 of the Guidelines on the scope of the obligations for general-purpose AI models.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI Act recognises that the entity that trains the model may not be the one that is contractually, economically, and/or technically in charge of the development (see the provider definition stating ‘has developed’ in Article 3(3) AI Act).

After
Removed.
removed
Untitled section
Before
The question of under whose authority a model is trained requires a case-by-case assessment, which can take account of contractual relations amongst other factors.

After
Removed.
removed
Untitled section
Before
Some examples are provided in Section 3.1.1 of the Guidelines on the scope of the obligations for general-purpose AI models.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Providers of general-purpose AI models must (Article 53 AI Act):

After
Removed.
removed
Untitled section
Before
- draw up and maintain technical documentation about the model, including information about the development process, to provide to the AI Office upon request; national competent authorities can also ask the AI Office to request information on their behalf when this information is necessary for them to exercise their supervisory tasks;

After
Removed.
removed
Untitled section
Before
- provide information and documentation to downstream AI system providers to help them understand the model's capabilities and limitations and comply with their own obligations;

After
Removed.
removed
Untitled section
Before
- implement a policy to comply with Union copyright law and related rights, including identifying and respecting rights reservations through state-of-the-art technologies; publish a sufficiently detailed summary of the content used for training the model; if they are established outside the EU, appoint an authorised representative in the Union before placing their model on the market.

After
Removed.
removed
Untitled section
Before
Providers can demonstrate compliance through the General-Purpose AI Code of Practice which was assessed as adequate, or via alternative adequate means.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
In addition to the standard obligations for providers of all general-purpose AI models, providers of general-purpose AI models with systemic risk must (Article 55 AI Act):

After
Removed.
removed
Untitled section
Before
- perform model evaluation using standardised protocols and state-of-the-art tools, including conducting and documenting adversarial testing to identify and mitigate systemic risks;

After
Removed.
removed
Untitled section
Before
- assess and mitigate possible systemic risks at Union level, including their sources, that may stem from the development, placing on market, or use of these models;

After
Removed.
removed
Untitled section
Before
- track, document, and report relevant information about serious incidents and possible corrective measures to the AI Office and, as appropriate, national authorities without undue delay;

After
Removed.
removed
Untitled section
Before
- ensure adequate cybersecurity protection for both the model and its physical infrastructure, to prevent unauthorised access, theft, or leakage.

After
Removed.
removed
Untitled section
Before
Providers of such models can demonstrate compliance through adhering to the General-Purpose AI Code of Practice or show alternative adequate means of compliance.

After
Removed.
removed
Untitled section
Before
If providers choose to comply via alternative means, they must present arguments for why such means are adequate, for assessment by the European Commission.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Providers of general-purpose AI models released as open-source may be exempt from certain obligations (Articles 53(2), 54(6) AI Act), specifically:

After
Removed.
removed
Untitled section
Before
- the requirement to maintain technical documentation for authorities;

After
Removed.
removed
Untitled section
Before
- the requirement to provide documentation to downstream AI system providers;

After
Removed.
removed
Untitled section
Before
- the requirement to appoint an authorised representative (for non-EU providers).

After
Removed.
removed
Untitled section
Before
These exemptions require that the general-purpose AI model:

After
Removed.
removed
Untitled section
Before
- is released under a free and open-source license, allowing access, use, modification, and distribution without monetisation;

After
Removed.
removed
Untitled section
Before
- has its parameters, including weights, architecture, and usage information publicly available;

After
Removed.
removed
Untitled section
Before
- is not a general-purpose AI model with systemic risk — providers of general-purpose AI models with systemic risk must comply with all the obligations for providers of general-purpose AI models, regardless of whether the model is released as open-source.

After
Removed.
removed
Untitled section
Before
These exemptions recognise that open-source models contribute to research and innovation while already providing transparency through their open nature.

After
Removed.
removed
Untitled section
Before
Nevertheless, providers whose models meet the above open-source requirements are not exempt from the copyright policy obligation or the requirement to publish a training data summary (Article 53(1)(c)(d) AI Act), since their open-source nature does not necessarily make available information on the data used for training or modifying the model, nor on how compliance with copyright law was ensured.

After
Removed.
removed
Untitled section
Before
Section 4 of the European Commission Guidelines on the scope of the obligations for general-purpose AI models provides further guidance on the open-source exemption.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The General-Purpose AI Code of Practice is a voluntary tool, prepared by independent experts in a multi-stakeholder process, designed to help industry comply with the AI Act’s obligations for providers of general-purpose AI models, ensuring that general-purpose AI models placed on the European market are safe and transparent, including the most powerful ones.

After
Removed.
removed
Untitled section
Before
It outlines a way for providers of general-purpose AI models and of general-purpose AI models with systemic risk to demonstrate compliance with the AI Act’s obligations laid down in Articles 53 and 55 AI Act.

After
Removed.
removed
Untitled section
Before
The Code has three chapters: Transparency and Copyright, both addressing all providers of general-purpose AI models, and Safety and Security, relevant only to the limited number of providers of the most advanced models that are subject to the AI Act's obligations for providers of general-purpose AI models with systemic risk.

After
Removed.
removed
Untitled section
Before
Any provider (or potential future provider) of a general-purpose AI model can sign the code by completing the Signatory Form and sending it to EU-AIOFFICE-CODE-SIGNATURES@ec.europa.eu.

After
Removed.
removed
Untitled section
Before
The form should be signed by a person with sufficient authority to bind the provider to the Code (e.g. a senior executive).

After
Removed.
removed
Untitled section
Before
Further information about the signing process can be found in these Q&A.

After
Removed.
removed
Untitled section
Before
For further information on which AI models are ‘general-purpose’ and under which conditions they are classified as ‘systemic risk’, see the other questions and Section 2 of the Guidelines on the scope of the obligations for general-purpose AI models.

After
Removed.
removed
Untitled section
Before
To learn more, read about the General-Purpose AI Code of Practice in these Q&A.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The European Commission will regularly monitor and evaluate the achievement of the objectives of the General-Purpose AI Code of Practice (Code).

After
Removed.
removed
Untitled section
Before
The updating mechanism is further elaborated in the European Commission’s adequacy assessment of the Code.

After
Removed.
removed
Untitled section
Before
In particular, the European Commission will consider facilitating formal updates to the Code at least every two years, for instance based on the emergence of standards, relevant technological developments, or changes in the risk landscape.

After
Removed.
removed
Untitled section
Before
To monitor the achievement of the objectives of the Code, the European Commission will remain in exchange with the signatories to understand where implementation support is necessary and may cooperate with national competent authorities, downstream providers, rightsholders, and other actors.

After
Removed.
removed
Untitled section
Before
Further, the Code leaves the European Commission’s responsibility unaffected to issue guidance on the application of the AI Act, which may be of relevance for concepts in the Code.

After
Removed.
removed
Untitled section
Before
In particular, in the case of an imminent threat of large-scale irreversible harm or to address its negative effects, the European Commission will consider whether rapid guidance on the AI Act’s application, or a rapid update to the Code agreed upon by the signatories, are appropriate, in addition to adequate enforcement actions.

After
Removed.
removed
Untitled section
Before
The Guidelines on the scope of the obligations for general-purpose AI models will be regularly reviewed and updated as appropriate.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Pursuant to Article 53(1)(a) and Annex XI, Section 1, point 2(e), of the AI Act, providers of general-purpose AI models must document the known or estimated energy consumption of their model.

After
Removed.
removed
Untitled section
Before
If it is unknown, this estimation may be based on the computational resources used.

After
Removed.
removed
Untitled section
Before
Moreover, the European Commission is empowered to adopt a delegated act to detail measurement and calculation methodologies that providers should use to measure or estimate the energy consumption of their models, to allow for comparable and verifiable documentation.

After
Removed.
removed
Untitled section
Before
For the period until such a delegated act is adopted, the Model Documentation Form in conjunction with the Transparency Chapter of the General-Purpose AI Code of Practice (Code) provides guidance on how providers may demonstrate compliance with this requirement, under ‘Energy consumption (during training and inference)’.

After
Removed.
removed
Untitled section
Before
In particular, if a Signatory to the Code does not know the energy consumption of training their model, they commit to reporting an estimated amount unless they are lacking critical information about compute or hardware which prevents them from being able to make an estimate.

After
Removed.
removed
Untitled section
Before
In this case, the Signatory commits to documenting the information they lack.

After
Removed.
removed
Untitled section
Before
For the purpose of estimating energy consumption, the AI Office will use the information available about the computational resources used for training to derive an estimate.

After
Removed.
removed
Untitled section
Before
To do so, the AI Office will rely on knowledge of what critical information the provider is lacking, available scientific resources, as well as on preliminary results from the ongoing European Commission study on energy-efficient and low-emission AI and draw on expertise from the scientific panel.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Providers of general-purpose AI models with systemic risk need to keep track of, document, and report, without undue delay, to the AI Office and, as appropriate, to national competent authorities, relevant information about serious incidents and possible corrective measures to address them (Article 55(1)(c) AI Act).

After
Removed.
removed
Untitled section
Before
The European Commission considers that this obligation covers serious cybersecurity breaches related to the model or its physical infrastructure, including the (self-)exfiltration of model parameters and cyberattacks, due to their possible implications for the obligations provided for in Article 55(1)(b) and (d) AI Act.

After
Removed.
removed
Untitled section
Before
Apart from this, the European Commission considers a ‘serious incident’ in the context of Chapter V AI Act as any incident or malfunctioning of a general-purpose AI model that directly or indirectly leads to any of the events listed in the corresponding definition for AI systems in Article 3(49)(a) to (d) AI Act.

After
Removed.
removed
Untitled section
Before
The Safety and Security Chapter of the General-Purpose AI Code of Practice, by way of its Commitment 9, provides a means to demonstrate compliance with this obligation.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The European Commission expects providers to comply with the obligations for providers of general-purpose AI models in the AI Act, which entered into application in August 2025.

After
Removed.
removed
Untitled section
Before
Some of these are straightforward — for instance, the notification to the AI Office when a provider is training a general-purpose AI model with systemic risk in Article 52 AI Act.

After
Removed.
removed
Untitled section
Before
Providers are expected to send these notifications without any further delay.

After
Removed.
removed
Untitled section
Before
Other obligations are more complex.

After
Removed.
removed
Untitled section
Before
For those, providers may demonstrate compliance through the General-Purpose AI Code of Practice (Code) or by demonstrating alternative adequate means of compliance.

After
Removed.
removed
Untitled section
Before
The European Commission understands that, especially regarding the more complex obligations, it may not be possible for some providers to fully implement the measures in the Code, or similarly adequate means of demonstrating compliance, by August 2025.

After
Removed.
removed
Untitled section
Before
In cases where a provider is not fully compliant with all the Code’s commitments right away, the European Commission will consider them to act in good faith and will be ready to collaborate to find ways to ensure full compliance.

After
Removed.
removed
Untitled section
Before
In particular, the European Commission encourages close informal cooperation with providers during the training of their general-purpose AI models with systemic risk to facilitate compliance and ensure timely market placement.

After
Removed.
removed
Untitled section
Before
Furthermore, the European Commission expects proactive reporting by providers of general-purpose AI models with systemic risk, whether as part of commitments under the Code or as part of alternative means of demonstrating compliance.

After
Removed.
removed
Untitled section
Before
This is particularly important during the first year, where the European Commission will focus on technical exchanges on working level to facilitate providers’ full compliance as soon as possible.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
AI regulatory sandboxes are safe and controlled environments for the experimentation, development, training and testing of innovative AI systems.

After
Removed.
removed
Untitled section
Before
Competent authorities supervising the sandboxes can provide guidance and advice on the interpretation of provisions of the AI Act and other relevant Union or national legislation.

After
Removed.
removed
Untitled section
Before
Any provider with an AI system that falls within the scope of the AI Act can apply for participation in an AI regulatory sandbox.

After
Removed.
removed
Untitled section
Before
Member States should have their AI regulatory sandboxes in place by 2 August 2026, after which (prospective) providers can apply for participation in the sandboxes.

After
Removed.
removed
Untitled section
Before
At least one AI regulatory sandbox will be available in each Member State.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Member States should have in place an AI regulatory sandbox with national coverage by 2 August 2026.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
Any provider with an AI system that falls within the scope of the AI Act can apply for participation in an AI regulatory sandbox.

After
Removed.
removed
Untitled section
Before
This means that AI regulatory sandboxes can (among others) supervise specific AI projects to fulfil the requirements and obligations in the AI Act, including for conformity assessment of high-risk AI systems, compliance of interactive and generative AI systems with the transparency requirements in Article 50 of the AI Act, or to enable identification, testing and implementation of effective preventive and mitigating measures to ensure specific borderline cases of AI systems do not constitute prohibited practices.

After
Removed.
removed
Untitled section
Before
Not in the old version.

After
Removed.
removed
Untitled section
Before
The AI regulatory sandboxes and the AI Act Service Desk are expected to complement each other’s aim.

After
Removed.
removed
Untitled section
Before
The AI regulatory sandboxes are designed for more specific and complex regulatory questions and challenges, and the aim of the AI Act Service Desk is to provide answers to more general questions concerning the AI Act.

After
Removed.
removed
Untitled section
Before
The regulatory challenges dealt with, and questions answered in the AI regulatory sandboxes can also support the aim of the AI Act Service Desk by making use of the lessons learnt, the guidance and answers provided in the AI regulatory sandboxes.

After
Removed.
added
- Digital Omnibus
Before
Not in the old version.

After
- Digital Omnibus
added
- Digital Omnibus
Before
Not in the old version.

After
Why is the European Commission proposing to amend the AI Act?
added
- Digital Omnibus
Before
Not in the old version.

After
The AI Act entered into force on 1 August 2024.
added
- Digital Omnibus
Before
Not in the old version.

After
It follows a staggered entry into application, with some parts already applicable such as certain prohibitions, AI literacy, and rules for general-purpose AI models.
added
- Digital Omnibus
Before
Not in the old version.

After
Other parts of the Act are set to apply on 2 August 2026 and 2 August 2027.
added
- Digital Omnibus
Before
Not in the old version.

After
This progressive roll-out allows us to build on the experience gathered in applying the first part of the rules.
added
- Digital Omnibus
Before
Not in the old version.

After
The Commission is committed to continuously learn and stepping up its efforts.
added
- Digital Omnibus
Before
Not in the old version.

After
This is particularly important in the context of a fast-evolving technology like AI.
added
- Digital Omnibus
Before
Not in the old version.

After
Stakeholder consultations throughout 2025 revealed implementation challenges that need to be addressed so that the AI Act can be successfully rolled-out.
added
- Digital Omnibus
Before
Not in the old version.

After
This proposal puts forwards legislative amendments to that effect and complements ongoing efforts to facilitate compliance with the AI Act, like the launch of an AI Act Service Desk.
added
- Governance & Enforcement
Before
Not in the old version.

After
- Governance & Enforcement
added
- Governance & Enforcement
Before
Not in the old version.

After
What is enforcement, and why does it matter?
added
- Governance & Enforcement
Before
Not in the old version.

After
Enforcement refers to the set of powers and procedures used to ensure that the AI Act is applied in practice, effectively supervised, and complied with by the operators concerned.
added
- Governance & Enforcement
Before
Not in the old version.

After
The AI Act’s enforcement framework turns the rules into practical action.
added
- Governance & Enforcement
Before
Not in the old version.

After
It helps make sure that operators developing or using AI in the EU follow the law in a consistent and effective way, with supervision and corrective powers where needed.
added
- Governance & Enforcement
Before
Not in the old version.

After
In practice, enforcement includes:
added
- monitoring compliance with the AI Act,
Before
Not in the old version.

After
- monitoring compliance with the AI Act,
added
- gathering the information needed to assess whether rules are being followed,
Before
Not in the old version.

After
- gathering the information needed to assess whether rules are being followed,
added
- gathering the information needed to assess whether rules are being followed,
Before
Not in the old version.

After
- requiring corrective action where providers of GPAI models or AI systems do not comply with the rules,
added
- gathering the information needed to assess whether rules are being followed,
Before
Not in the old version.

After
- using stronger corrective measures where necessary, including restrictions to the availability of GPAI models and AI systems.
added
- Governance & Enforcement
Before
Not in the old version.

After
- Governance & Enforcement
added
- Governance & Enforcement
Before
Not in the old version.

After
What are the entry channels for complaints?
added
- Complaints Tool for AI Systems
Before
Not in the old version.

After
- Complaints Tool for AI Systems
added
- Complaints Tool for AI Systems
Before
Not in the old version.

After
The AI Act Complaints Tool for AI Systems is a channel made available by the AI Office under Article 85 of the AI Act for natural and legal persons who wish to submit a complaint concerning alleged infringements to the AI Act for AI systems falling within its exclusive competence.
added
- Complaints Tool for AI Systems
Before
Not in the old version.

After
Under Article 75(1) of the AI Act, the AI Office has exclusive competence for two main categories of AI systems:
added
- Complaints Tool for AI Systems
Before
Not in the old version.

After
- GPAI-based AI systems, where the general-purpose AI model and the AI system are provided by the same provider or a provider within the same undertaking as the provider of the underlying GPAI model.
added
- Complaints Tool for AI Systems
Before
Not in the old version.

After
This does not include systems covered by Annex I, Annex III point 2, or Annex III point 8 of the AI Act, as well as AI systems provided by law enforcement authorities, border management authorities and financial institutions, insofar as those AI systems fall under Article 74(6).
added
- Complaints Tool for AI Systems
Before
Not in the old version.

After
The AI systems excluded remain in the enforcement remit of national competent authorities.
added
- Complaints Tool for AI Systems
Before
Not in the old version.

After
- AI systems embedded in, or constituting, very large online search engines (VLOSEs) or very large online platforms (VLOPs) within the meaning of the Digital Services Act.
added
LINK: AI Act Complaints Tool for AI Systems
Before
Not in the old version.

After
LINK: AI Act Complaints Tool for AI Systems
added
- Whistleblower Tool
Before
Not in the old version.

After
- Whistleblower Tool
added
- Whistleblower Tool
Before
Not in the old version.

After
The AI Act Whistleblower Tool, empowers individuals who are professionally connected to an AI provider to securely submit a report and contribute directly to make AI in Europe safe transparent, and trustworthy.
added
- Whistleblower Tool
Before
Not in the old version.

After
Whistleblowers play a vital role in identifying potential violations of the law that could endanger fundamental rights, health, or public trust, and which might otherwise go undetected.
added
- Whistleblower Tool
Before
Not in the old version.

After
By reporting potential violations, whistleblowers can support the AI Office in detecting them early on, thereby contributing to the safe and transparent development of AI technologies.
added
- Whistleblower Tool
Before
Not in the old version.

After
The AI Act Whistleblower Tool allows relevant individuals to report harmful practices by providers of GPAI AI models and certain AI systems under the AI Office’s remit.
added
- Whistleblower Tool
Before
Not in the old version.

After
You can submit your report anonymously in any EU language together with supporting documents using the secure inbox in the tool.
added
- Whistleblower Tool
Before
Not in the old version.

After
The inbox also enables whistleblowers to stay up to date on the progress of their report, answer follow-up questions while remaining anonymous.
added
LINK: AI Act Whistleblower Tool
Before
Not in the old version.

After
LINK: AI Act Whistleblower Tool
added
- Downstream providers dedicated channel for complaints related to GPAI models
Before
Not in the old version.

After
- Downstream providers dedicated channel for complaints related to GPAI models
added
- Downstream providers dedicated channel for complaints related to GPAI models
Before
Not in the old version.

After
Under the AI Act, downstream providers of an AI system built on a general-purpose AI model may lodge a complaint with the Commission where they consider that a provider of a general-purpose AI model has infringed Articles 53 to 55 of the AI Act, which set out the relevant obligations for such models.
added
- Downstream providers dedicated channel for complaints related to GPAI models
Before
Not in the old version.

After
The Commission will examine the information received and all complaints will be dealt in a confidential manner.
added
- Downstream providers dedicated channel for complaints related to GPAI models
Before
Not in the old version.

After
However, submission of a complaint does not automatically lead to the opening of proceedings, as the Commission retains discretion in how it prioritises and pursues enforcement action.
added
- Downstream providers dedicated channel for complaints related to GPAI models
Before
Not in the old version.

After
Complaints concerning infringement of the AI Act by the providers or deployers of AI systems, should be addressed through the dedicated channel, which contains a voluntary complaint template.

Every changed clause was found by code and is explained or listed as minor; every quote is checked word for word against both versions. Explanations are written by AI.

Watched pages join the open-source monitor, which publishes every change it finds: see what it found so far.

What changed in Ai-act-service-desk Frequently Asked Questions AI Act Service Desk?

Untitled section

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- Governance & Enforcement

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

- General-purpose AI models

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

- Digital Omnibus

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section

Untitled section