Keep Updates on AI Act (AIA): EU AIA enters into force.

Within the EU AIA, it aims to promote responsible AI development and ensure the protection of citizens’ health, safety, and fundamental rights while providing developers and businesses with clear guidelines and reducing administrative and financial burdens. Recently developed, the Commission has initiated a consultation for a Code of Practice for General-Purpose AI (GPAI) models, focusing on transparency, copyright, and risk management. Feedback will inform the code, expected to be finalized by April 2025, and assist the AI Office in overseeing its implementation. The framework and risk categories consist of four classes: minimal risk, specific transparency risk, high risk, and unacceptable risk [fC24, PtcotEU24]. The details are shown in Figure 1.

Diagram illustrating the AI risk classification under the EU AI Act. The categories include: 'Minimal Risk' in green, covering AI systems like spam filters and video games that don’t require strict standards; 'Specific Transparency Risk' in blue, for AI systems like chatbots that must identify as AI-generated; 'High Risk' in red, for AI requiring adherence to standards such as medical and recruitment tools; and 'Unacceptable Risk' in gray, prohibiting AI that violates fundamental rights, such as social scoring and certain government systems.
Fig 1. AI risk classifications

After being proposed by the European Artificial Intelligence Act (EU AIA) in April 2021 and agreed upon in December 2023. On August 1, 2024, the EU AIA enters into force across all 27 EU Member States; as a result, the majority of its rules will begin to be enforced on August 2, 2026. There are three points to provide after August 2, 2026 [Act24c] as:

• to apply chapters I and II from February 2, 2025.
• to apply chapters III, section 4, V, VII, XII, and Art. 78 from August 2, 2025 (exception of Art. 101).
• to apply Art. 6(1) and the corresponding obligations in this regulation from August 2, 2027.

The key and enforcement points of the EU AIA were recently updated after August 2, 2026.

  • The first comprehensive legal framework for the regulation of AI systems across the EU was published in The EU Official Journal, namely ”Regulation (EU) 2021/1689,” on July 12, 2024 [PtcotEU24].

  • The EU AIA enters into force on August 1, 2026, and will be effective from 2 August 2026 (Art. 113 EU AI Act), except for the specific provisions listed in Art. 113(a)-(c) EU AI Act:

    (a) the enforcement of Chapter I and II stats from February 2, 2025 (Art. 113(a) EU AIA).
    Description: it includes the general provisions, definitions, and rules regarding prohibited uses of AI.
    (b) the enforcement of certain requirements stats from August 2, 2025 (Art. 113(b) EU AIA). For the GPAI models placed on the EU market before August 2, 2025, achieve compliance until August 2, 2027 (Art. 111(3) EU AIA).
    Description: it includes the notification obligations, governance, rules on GPAI models, confidentiality, and penalties (other than penalties for providers of GPAI models).
    (c) the enforcement of Art. 6 stats from August 2, 2027 (Art. 113(c) EU AIA).
    Description: it includes the corresponding obligations regarding high-risk AI systems.

  • The General-Purpose Artificial Intelligence (GPAI) model placed on the EU market a) before August 2, 2025; enforcement begins on August 2, 2027 (Art. 111(3) EU AIA); and b) the GPAI model placed on the EU market on or after August 2, 2025; enforcement begins when the GPAI model is placed on the EU market (Art. 113(b) EU AIA).

  • Any high-risk AI systems are not intended for use by public authorities and placed on the EU market before August 2, 2026; there will be no enforcement of the rules on high-risk AI systems as long as such AI systems are not subject to any significant modifications (Art. 111(2)).

  • The high-risk AI systems under Art. 6(1) are placed on the EU market or subject to any significant modifications after August 2, 2026; enforcement of the rules on high-risk AI systems begins on August 2, 2027 (Arts. 6(1), 111(2), and 113(c)).
    Note: Art. 6(1) refers to AI systems that include safety components of the product or the product itself and are covered under the Union harmonization legislation listed in Section B of Annex I [Act24a].

  • The high-risk AI systems under Art. 6(2) are placed on the EU market or subject to any significant modifications after August 2, 2026; enforcement of the rules on high-risk AI systems begins when the AI system is placed on the EU market (Arts. 6(2), 111(2), and 113).
    Note: Art. 6(2) refers to AI systems that include biometrics, critical infrastructure, education, and vocational training, employment-worker’s management and access to self-employment, access to and enjoyment of essential private services and essential public services and benefits, law enforcement, migration, and administration of justice and democratic processes [Act24b].

  • Deepfakes. It means KI-generated or manipulated image, audio, or video content that resembles existing persons, objects, places, entities, or events and would falsely appear to a person to be authentic or truthful (Art. 3(60) EU AI Act). Under the EU AIA, the deepfakes created by AI systems are required to clearly describe that the content has been manipulated or artificially created by labeling the AI output as such and revealing its artificial origin.

    Note: the content forms part of an evidently artistic, creative, satirical, fictional, or analogous work or program; the transparency obligations set out in this paragraph are limited to disclosure of the existence of such generated or manipulated content in an appropriate manner that does not hamper the display or enjoyment of the work (Art. 50(4) EU AIA).

Point of requirement from member states


  • to create a minimum of one functional regulatory sandbox nationwide by August 2, 2026 (Art. 57(1) EU AIA).

  • to assign a minimum of one notification authority and one market monitoring authority by August 2, 2025 (Art. 70(2) EU AIA).

  • to inform the Commission of the primary point of contact and the identities of the appropriate authorities by August 2, 2025 (Art. 70(2) EU AIA).

  • to provide information to the public on how to get in contact with the appropriate authorities and one point of contact by August 2, 2025 (Art. 70(2) EU AIA).

Penalties


According to Art. 99(3) of the EU AIA, the maximum punishment for violating the regulations on the forbidden uses of AI is either a fine of up to EUR 35 million or 7 percent of global yearly revenue. Additionally, under Art. 85 to 87 of the EU AIA, natural and legal persons have the right to report instances of non-compliance, explain their decision-making, and file a complaint with a market monitoring. Also under Art. 99(1) of the EU AIA, member states must consider the economic viability and interests of SMEs, particularly start-ups, when determining the appropriate penalty levels for infractions [TH24].

References

[Act24a]         EU Artficial Intelligence Act. Annex i: List of union harmonisation legislation, 2024.

[Act24b]         EU Artficial Intelligence Act. Annex iii: High-risk ai systems referred to in article 6(2), 2024.

[Act24c]          EU Artficial Intelligence Act. Article 113: Entry into force and application, 2024. Directorate-General for Communication. Ai act enters into force, 2024.

[fC24]              Directorate-General for Communication. Ai act enters into force, 2024.

[PtcotEU24]   The European Parliament and the council of the European Union. Regulation (eu) 2024/1689 of the european parliament and the council, 2024.

[TH24]            Constantin Teetzmann, Aishwarya Jha, Tim Hickman, Sylvia Lorenz. The long-awaited EU AI act will become law after publication in the EU’s official journal in 2024.

 

 

 

Scroll to Top