EU AI Act Takes Full Effect: Fines of Up to €35 Million for High-Risk AI Systems

On 2 August 2026, the EU AI Act, the world's most sweeping law on artificial intelligence, took full effect across the European Union. From that day, companies using high-risk AI systems in the EU must meet the regulator's requirements or pay fines of up to €35 million.

What requirements are now in force?

As of 2 August 2026, all companies with high-risk AI systems on the EU market must complete full compliance. The law classes AI systems in healthcare, education, banking, hiring and critical infrastructure management as high-risk.

Key requirements:

• Mandatory risk assessment before a system enters the EU market
• Documentation of training data, architecture and algorithms
• Human oversight mechanisms and the ability to shut the system down
• Registration in the EU AI Office database
• Transparency: users must know they are interacting with AI

Fines reach up to €35 million or 7% of a company's global annual turnover, whichever is higher. Misleading the regulator carries a separate penalty of up to €7.5 million or 1% of turnover.

How the law was phased in

The European Parliament adopted the EU AI Act in March 2024, and it was officially published in July 2024. Its requirements were spread over two years to give companies time to adapt their products.

On 2 February 2025, bans on the most dangerous practices took effect: social scoring of citizens, manipulative AI that exploits subconscious influence, and unauthorized real-time biometric identification in public places. In August 2025, rules for developers of powerful GPAI models kicked in, requiring technical safety reports and systemic risk assessments.

"The EU AI Act is the world's first law to set legally binding requirements for AI.

It is a historic decision in favor of trustworthy, human-centric artificial intelligence," according to official European Commission materials.

What this means

The EU AI Act applies to every company that sells or uses AI in the EU, regardless of where it is registered. OpenAI, Google, Microsoft, and Asian and Russian developers must comply if their products are available to European users.

Analysts call this the "Brussels effect": the world's largest single market sets a standard that global companies often apply across all their markets. That is how the GDPR worked in 2018, when its personal data protection requirements became the de facto global standard. The EU AI Act aims to play the same role for AI regulation.

Frequently asked questions

Which AI systems are considered high-risk?

High-risk systems include AI used in medical diagnostics, educational assessment, credit scoring, candidate selection and critical infrastructure management, such as power grids, water supply and transport. The full list is set out in Annex III of the EU AI Act.

Which practices are banned outright?

Since February 2025, the EU has banned the following outright: social scoring of citizens, AI that uses manipulative subliminal techniques, and real-time biometric identification in public places (with strictly limited exceptions for law enforcement operations). Fines for these violations are the maximum: up to €35 million or 7% of turnover.