AI rules in Europe change on 2 August 2026. And there's less time than you think.

The EU AI Act is no longer a draft or a proposal. It's law. And its most critical date is less than four months away: on 2 August 2026, obligations for high-risk AI systems come fully into force.

There's a lot of confusion about what this actually means. Let's go through it clearly.

What's already in force

Since February 2025, systems considered to pose unacceptable risk are already banned: social scoring, subliminal manipulation and mass biometric surveillance in public spaces. Also mandatory since then: AI literacy training for all staff. Not in August. Now. Already.

What comes in August 2026

This is where things get serious for most organisations:

• High-risk AI systems must achieve full compliance: risk management, human oversight, data governance, technical documentation, conformity assessment, and registration in the EU database
• Article 50 transparency obligations come into force: if your system interacts with people, generates synthetic content or uses emotion recognition, you must disclose this
• AI agents are included. The European Commission confirmed this recently: there is no separate category — the same regulation applies to them

Which sectors are most affected

If you use AI for recruitment, credit scoring, biometric identification, critical infrastructure or education, your system is probably high-risk. Fines go up to €35 million or 7% of global turnover.

What your organisation should be doing right now

1. Inventory all AI systems you use — including those your employees use independently
2. Classify them by risk level according to Annex III of the Act
3. Begin implementation: documentation, conformity assessments, team training
4. Don't wait for the Digital Omnibus package — it's a Commission proposal, not an approved law, and there's no guarantee it will be adopted as drafted

How we can help

At AP Interactive we help organisations implement AI safely and in compliance with regulation: private models on own infrastructure, system auditing, data governance and real regulatory compliance.

Regulation isn't a threat. It's the opportunity to do this properly before you're forced to.

Frequently asked questions

What happens on 2 August 2026?

High-risk AI systems must achieve full compliance under the EU AI Act: risk management, human oversight, data governance, technical documentation, conformity assessment, and registration in the EU database. Article 50 transparency obligations also come into force, requiring disclosure whenever a system interacts with people, generates synthetic content, or uses emotion recognition.

Are AI agents covered by the EU AI Act?

Yes. The European Commission has confirmed there is no separate category for AI agents — the same regulation that applies to other AI systems applies to them. If an agent is used in a high-risk context such as recruitment or credit scoring, it must meet the same compliance requirements.

Should organisations wait for the Digital Omnibus package before acting?

No. The Digital Omnibus is a Commission proposal, not an approved law, and there is no guarantee it will be adopted as drafted. Organisations should inventory their AI systems, classify them by risk level under Annex III, and begin implementation now rather than waiting on a proposal that may change.

Get in touch to discuss your current AI exposure and compliance roadmap.