The EU AI Act started enforcing fines on August 2. Only 8 of 27 member states designated contacts.

The EU AI Act crossed from drafting into enforcement on August 2, 2026. Article 50 transparency obligations now apply. The AI Office can fine general-purpose AI providers up to 3% of global turnover.

Only 8 of 27 member states designated enforcement contacts before the deadline.

What August 2 activated

Article 50 of the EU AI Act requires transparency obligations for general-purpose AI (GPAI) models. Every model provider shipping into Europe must now comply. The fines reach up to 3% of global turnover.

The grace period closes December 2, 2026. That gives providers four months to reach full compliance.

For US and Chinese AI companies, the rules apply extraterritorially. If your model is accessible in the EU, you are subject to the regulation regardless of where your company is incorporated.

The enforcement gap

Only 8 of 27 member states designated enforcement contacts with the AI Office before the August 2 deadline. That means the regulation is active, but most of the bloc cannot yet enforce it.

The AI Office in Brussels has the authority to fine GPAI providers. But the member states are the ones who handle investigations, audits, and day-to-day compliance checks. Without designated contacts in 19 countries, the enforcement mechanism has a structural gap.

"Regulators stopped treating AI regulation 2026 as a future agenda item and started issuing fines, audit letters, and procurement checklists," according to analysis from Beyond TMRW.

The EU AI Act enforcement calendar entered its operational phase for GPAI models in 2026. More than a dozen US state AI laws also took effect or advanced in legislatures from Colorado to California.

What providers must do

The Article 50 obligations include:

• Transparency about AI-generated content
• Documentation of training data
• Disclosure of model capabilities and limitations
• Compliance with copyright obligations for training data

Anthropic signed the Article 50(2) Code of Practice on Transparency of AI-Generated Content. Google already watermarks with SynthID. OpenAI signed the same Code of Practice and built a watermark implementation in 2024.

The compliance infrastructure is building from the top down. The regulation is enforcing from the bottom up.

The December 2 deadline

The grace period closes December 2, 2026. After that date, the AI Office can begin issuing fines for non-compliance.

The fine structure is significant. Up to 3% of global turnover for GPAI providers means the largest AI companies face potential fines in the billions. For a company with $40 billion in revenue, a 3% fine reaches $1.2 billion.

The enforcement is not theoretical. The EU has a track record of using fines as a compliance mechanism. GDPR fines reached billions of dollars against major technology companies. The AI Act fine structure follows the same playbook.

What this means for the market

The EU AI Act creates a compliance floor that every model provider must meet. The companies that reach compliance first gain a market advantage.

The regulation also creates a market for compliance tooling. Companies that help providers document their training data, implement transparency measures, and manage copyright obligations are positioning for a market that did not exist before August 2.

For enterprise buyers, the regulation creates procurement requirements. If you deploy AI models in Europe, your vendors must comply with Article 50. That means your procurement process now includes a compliance check that did not exist six months ago.

Source