EU AI Act Takes Effect: New Rules and Implications for AI Providers and Users

EU AI Act: A Landmark Regulation

The European Union's Artificial Intelligence Act (EU AI Act), described as "the world's first comprehensive AI law," is progressively becoming a reality for the 450 million people living in the 27 EU countries 1

1

. As of August 2, 2025, significant provisions of the Act will come into force, particularly affecting providers of General Purpose AI (GPAI) models 2

2

3

3

.

Key Provisions and Deadlines

The EU AI Act adopts a risk-based approach, categorizing AI systems based on their potential risks and impact on individuals and society 5

5

. Key provisions include:

1. National Oversight: EU member states must notify the European Commission about appointed market surveillance authorities to oversee businesses' compliance with the AI Act 3

   3

   .

2. Penalties: Companies may face fines of up to €35 million or 7% of total worldwide annual turnover for breaches of the AI Act, whichever is higher 3

   3

   5

   5

   .

3. GPAI Regulations: Rules on general purpose AI systems, including large language models, will enter into force 3

   3

   5

   5

   .

Implications for AI Providers and Users

GPAI model providers must adhere to specific stipulations, including:

1. Maintaining current technical documentation and training data summaries 5

   5

   .
2. Implementing a copyright compliance policy 4

   4

   .
3. Providing guidance for downstream deployers 5

   5

   .
4. Ensuring transparency regarding capabilities, limitations, and intended use 5

   5

   .

For GPAI models deemed to pose "systemic risk," additional requirements include conducting model evaluations, reporting incidents, implementing risk mitigation strategies, and disclosing energy usage 5

5

.

Global Impact and Industry Response

The EU AI Act isn't just a regional regulation; it's set to influence AI governance, risk management, and compliance practices worldwide 2

2

. Major tech companies have responded differently to the Act's voluntary Code of Practice for GPAI Providers:

• Google, OpenAI, and others have signed the code 1

  1

  4

  4

  .
• Meta has declined to sign, citing concerns about stifling innovation 3

  3

  .

Challenges and Criticisms

The implementation of the EU AI Act faces several challenges:

1. Many member states may miss the August 2 deadline to appoint national regulators 3

   3

   .
2. There are concerns about the complexity of coordinating multiple bodies at both EU and national levels 3

   3

   .
3. Some copyright holders express dissatisfaction with the GPAI Code, claiming it doesn't provide meaningful protection of intellectual property 4

   4

   .

Preparing for Compliance

Companies operating in or serving the EU market must prepare for these new regulations:

1. Understand the risk categories and requirements applicable to their AI systems 1

   1

   2

   2

   .
2. Implement necessary documentation and transparency measures 5

   5

   .
3. Engage with appointed Notified Bodies for conformity assessments of high-risk AI systems 5

   5

   .
4. Prepare for potential audits and oversight from national authorities 3

   3

   .

As the first binding set of rules on AI with global reach, the EU AI Act represents a significant step towards trustworthy AI and responsible innovation 4

4

. While not perfect, it offers practical steps toward building more responsible AI systems, including stronger data governance, privacy, security, and risk oversight 2

2

.