Key Provisions of the EU AI Act

European Union officials have reached a historic provisional agreement on the world’s first comprehensive AI Act. The landmark legislation aims to regulate artificial intelligence based on its potential risks and impacts. This sets a global benchmark for AI governance.

The political deal was struck after lengthy negotiations. It establishes a legal framework for AI development and use. According to Reuters, the rules are designed to foster innovation while protecting fundamental rights.

The new law takes a risk-based approach to artificial intelligence. It outright bans AI systems deemed to pose an unacceptable risk. These include manipulative techniques and social scoring by governments.

High-risk AI applications will face strict obligations. This includes mandatory fundamental rights impact assessments. Sectors like critical infrastructure and medical devices fall into this category.

The Act also introduces specific rules for general-purpose AI models. Developers of powerful models must conduct evaluations and assess systemic risks. They must also report any serious incidents to the European Commission.

Transparency requirements are a major component of the new rules. Citizens will have a right to know when they are interacting with an AI system. Deepfakes and AI-generated content must be clearly labeled.

Global Impact and Implementation Timeline

The EU AI Act is expected to have a significant global influence. It could become a worldwide standard, similar to the GDPR data law. Companies outside the EU must comply if they want to operate in the bloc’s single market.

Enforcement will be carried out by a new European AI Office. This body will supervise general-purpose AI models. National authorities will monitor the implementation at a local level.

Violations of the law could lead to substantial fines. Companies may face penalties of up to 35 million euros or 7% of global turnover. The rules are expected to fully apply by 2026, with some bans effective sooner.

The EU AI Act establishes a pioneering path for responsible technology development. It balances innovation with crucial safeguards. This framework will likely shape the future of artificial intelligence worldwide.

Info at your fingertips

When will the EU AI Act come into force?

The Act is expected to be fully applicable by 2026. Some provisions, including bans on certain AI practices, could take effect sooner. The timeline allows for a gradual implementation.

What AI systems are banned under the new law?

The law bans AI for social scoring and manipulative techniques. It also prohibits emotion recognition in workplaces and schools. Predictive policing based solely on profiling is forbidden.

How will the Act affect companies outside Europe?

Any company offering AI systems in the EU market must comply. This includes foreign developers and service providers. The rules have an extraterritorial scope, similar to the GDPR.

What are the penalties for non-compliance?

Fines can reach up to 35 million euros or 7% of global annual turnover. The amount depends on the infringement and the company’s size. Minor violations have lower fine structures.

Who will enforce the AI Act?

A new European AI Office will supervise general-purpose AI. National authorities in member states will enforce the rules locally. An AI Board will facilitate coordination and implementation.