What Is The EU AI Act?


The European Union is taking a leading role in regulating artificial intelligence to ensure it is trustworthy and protects fundamental rights. The proposed EU AI law aims to establish harmonized rules for the development, marketing and use of AI in Europe.

Overview of the EU AI law

The EU AI Law, proposed in April 2021, is landmark legislation that would provide the first comprehensive framework for AI regulation. The main objectives are:

  • Make Europe a global leader in ethical and trustworthy AI
  • Protect fundamental rights and security from high-risk AI systems
  • Remove legal barriers to AI innovation and investment
  • Improve governance and enforcement mechanisms

The law takes a risk-based approach to regulate AI in relation to the risks associated with it. It bans certain uses of AI that are deemed unacceptable. It places strict requirements on high-risk AI systems and limits obligations for lower-risk AI.

READ ALSO: I asked AI what Europeans think Americans from every single state look like

Key provisions of the AI ​​Act

The AI ​​Act establishes clear definitions and divides AI systems into four risk categories:

1. Unacceptable risks of AI systems

  • Completely prohibited with limited exceptions
  • Examples: AI for social scoring, mass surveillance, subliminal techniques

2. High-risk AI systems

  • Allowed, but subject to strict obligations
  • Must undergo conformity assessments before being placed on the market
  • Examples: AI in critical infrastructure, credit scoring, recruitment tools

3. Limited risk AI systems

  • Only subject to transparency obligations
  • Examples: AI chatbots, deepfakes

4. AI systems with minimal risk

  • Exempt from most obligations
  • Examples: AI in video games, spam filters

Key requirements for high-risk systems include:

  • High quality training, validation and test data
  • Documentation of datasets and programming
  • An appropriate risk management system
  • High level of algorithmic robustness, accuracy and cybersecurity
  • Clear functionalities and purpose limitations
  • Human supervision with the ability to overwrite
  • High degree of transparency for users
  • Logging capabilities and effective auditing

Who does the EU AI law apply to?

The EU AI law applies to providers, users, distributors, importers and manufacturers involved in AI systems that impact the EU market.

Providers refer to developers of AI systems or those who have developed AI systems to bring to the EU market.

Users refer to natural or legal persons who use AI systems in the context of a personal or professional activity.

The AI ​​law also applies extraterritorially. For example, an AI provider based in Switzerland must comply if it targets EU users.

READ ALSO: AI Summit Bletchley Park: the epicenter of AI safety

Governance under the EU AI Act

The EU AI Act establishes a European Artificial Intelligence Board to oversee implementation. It will consist of representatives of the EU Member States and aims to:

  • Advising and issuing opinions to national supervisory authorities
  • Promote coordination between national authorities
  • Support cross-border cooperation in enforcement
  • Maintaining contacts with industry and stakeholders
  • Follow developments in the field of AI
  • Contribute to guidance in the field of technical specifications

The board will help ensure uniform application of the AI ​​law across the EU.

Sanctions for non-compliance

High fines are provided for violating the EU AI Act:

  • For companies: up to €30 million or 6% of global turnover
  • For false or misleading information to authorities: up to €10 million or 2% of global turnover

These significant penalties demonstrate that the EU intends to seriously enforce the AI ​​law.

AI innovation and competitiveness

The EU aims to regulate the use of high-risk AI, but also wants to encourage the adoption and investment in AI. Provisions in the law aim to:

  • Promote small-scale providers and users through exemptions and lighter requirements
  • Provide easier access to data needed to train AI systems
  • Allow experimental and open-source AI development
  • Supporting international cooperation on standards

However, some companies have argued that the law could negatively impact the EU’s technological progress and competitiveness. The Commission says it aims to find the right balance between protection and innovation.

READ ALSO: What are the objectives of the AI ​​Safety Summit?

EU versus US approaches to AI regulation

The EU’s comprehensive regulatory approach differs from the sectoral, risk-based approach in the United States.

The main differences include:


  • Horizontal rules covering all sectors
  • Mandatory requirements for high-risk AI
  • Centralized administration


  • Vertical rules for individual sectors
  • Voluntary best practices
  • Decentralized approach

This contrast reflects philosophical disagreements over whether AI requires specific regulations covering all uses.

Criticism and challenges

There has been debate about some provisions of the law, including:

  • Broad definitions and risk categories
  • Feasibility for SMEs and startups
  • Costs of compliance and consequences for innovation
  • Enforceability and keeping up with AI developments

Critics argue that the law takes a precautionary approach, which risks stifling innovation. But advocates say building trust is necessary for long-term AI success.

The main challenges will be refining definitions, dealing with fragmentation between Member States and future-proofing regulations.

Next steps for the EU AI Act

The AI ​​law is still a proposal that is undergoing final negotiations between the EU institutions. Key next steps include:

  • The Parliament and the Council are negotiating a common position
  • Further changes possible
  • Final legislative act formally adopted
  • 2-4 years for access to the application
  • European AI Board established
  • National authorities have been designated to monitor implementation

The timeline remains uncertain, but the general direction has been set. The EU AI Act represents a global milestone for AI governance that will be closely watched worldwide.


The EU AI Act establishes groundbreaking regulations on artificial intelligence to ensure it develops responsibly. A nuanced approach is needed, tailored to different levels of risk. The compliance obligations for high-risk AI are significant and aim to uphold fundamental rights. Companies will need to proactively assess their AI practices and systems. With the right balance, the law can enable reliable AI innovation across Europe. But its implementation in the real world will require complex navigation between protection and progress to shape the future of AI.

Leave a Comment