The European Union has announced that the AI Act obligations for providers of general-purpose AI (GPAI) models come into effect from tomorrow.

This move aims to bring more transparency, safety, and accountability to AI systems across the EU market.

According to the European Commission, the new rules will ensure:

• Clearer information on how AI models are trained
• Better enforcement of copyright protections
• More responsible AI development

To support providers, the Commission has released guidelines clarifying who must comply with the new obligations. GPAI models are defined as those trained with over 10^23 FLOP and capable of generating language.

Additionally, a template has been published to help providers summarise the data used for model training.

The Commission, along with EU Member States, also revealed that the GPAI Code of Practice developed by independent experts serves as an effective voluntary tool for compliance.

Providers adhering to the Code will benefit from reduced regulatory burdens and increased legal clarity.

From August 2, 2025, providers must meet transparency and copyright requirements before placing GPAI models on the EU market.

Existing models launched before this date must ensure compliance by August 2, 2027.

Moreover, providers of advanced models presenting systemic risks those exceeding 10^25 FLOP will face additional obligations. These include mandatory notification to the Commission and ensuring enhanced safety and security standards.

Related post

View all