Questions about the AI Act?
We'll guide you safely through the new regulations.
Inquire without obligationFree & non-binding.
BREAKING NEWS
AI Act is Now Active:
What Applies as of TODAY
A new era for Artificial Intelligence in Europe has begun. The AI Act now forces AI developers into a tight framework of transparency and security. The key facts about the new legal situation.

1. The New Rules in Plain Terms: What Applies Now?
The deadline is here: Since August 2, 2025, providers of general-purpose AI models (GPAI) like OpenAI, Google, and Anthropic must comply with the new AI Act regulations. This is no longer a drill – the rules are binding law. At its core, there are three hard obligations:
Disclosure of Training Data
Companies must publish a detailed summary of the data used to train their models. This directly targets the trade secrets of AI forges and is intended to help copyright holders enforce their rights.
Proof of Copyright Compliance
Providers must present a functioning policy that proves how they respect and protect EU copyright law. Blanket claims are no longer sufficient.
Complete Technical Documentation
The entire technical development and capabilities of the model must be meticulously documented and made accessible to authorities.
2. Scandal at Launch: Meta on Collision Course
To help with implementation, the EU Commission has presented a voluntary code of conduct, the "Code of Practice". Those who sign it signal cooperation. 26 companies, including heavyweights like Google, Microsoft, and OpenAI, are on board.
But Meta, the company behind Llama, Facebook, and Instagram, refuses to sign. Vice President Joel Kaplan called the code an "overreaction" that slows innovation in Europe. This open confrontation could set the direction and shows the explosive force of the new regulation.
Elon Musk's xAI is also reluctant and has only accepted parts of the code. These companies must now prove in other, presumably more complex ways, that they comply with the laws.
3. Draconian Penalties Loom: What Happens with Violations
The EU is still granting one last deadline. Providers have exactly one year to fully adapt their systems. But from August 2, 2026, things get serious: The EU Commission can then impose fines.
And they're significant: For false information to authorities alone, penalties of up to 15 million euros or 3% of global annual turnover can be imposed. Violations of core obligations can be even more expensive.
Important for those affected: Civil lawsuits by citizens or competitors are now possible based on the new regulation, even if the Commission itself has not yet imposed sanctions.
Read Also
When Does What Apply? The Complete Timeline of the AI Regulation
All deadlines, dates, and transitional provisions of the EU AI Regulation in an interactive timeline.
Read articleEU AI Act: Commission Publishes Template for AI Training Data
The EU Commission has published an official template for documenting training data. What this means for providers.
Read article