AI Act Enforcement Is Live: What Providers Must Know from 2 August 2026

As of 2 August 2026, AI Act enforcement is live across the EU. The AI Office now has full power to investigate and fine providers of general-purpose AI models, and enforcement of the prohibited practices ban has begun.

Whether you are a GPAI provider, a deployer of AI systems, or a company incorporating AI into products, the compliance clock is no longer theoretical. Here is what you need to know.

1. AI Office Enforcement Powers

The AI Office can now:

• Request documentation and technical records
• Access general-purpose AI models for evaluation
• Impose fines of up to €15 million or 3% of global turnover

2. Prohibited Practices Carry the Highest Fines

Article 5 prohibitions are now in force. Banned practices include:

• Manipulation and subliminal techniques
• Exploiting vulnerabilities of individuals or groups
• Social scoring
• Predictive policing through profiling

Penalties reach €35 million or 7% of global turnover — the highest tier under the AI Act.

3. Transparency Rules Are Active, With a Short Grace Period

Article 50 transparency obligations apply from 2 August 2026. Systems already on the market have until 2 December 2026 to add machine-readable AI content marking.

If your system generates synthetic text, images, audio, or video, now is the time to audit your labelling pipeline.

4. High-Risk AI Systems Got a Delay, Not an Exemption

The Digital Omnibus pushed high-risk obligations back, but they are still coming:

• Standalone high-risk AI systems: 2 December 2027
• Embedded high-risk AI systems: 2 August 2028

This is a delay, not a reprieve. The groundwork — risk management, data governance, technical documentation, and human oversight — should start now.

Bottom Line

Enforcement has started for GPAI providers and prohibited practices. Transparency labelling has a short runway. High-risk obligations moved out, but preparation should begin immediately.