General-purpose and generative AI models

A general-purpose AI model placed on the EU market carries obligations of its own, separate from the risk classification of any system built on it. The provider must keep technical documentation, give downstream providers what they need to comply, adopt a copyright policy and publish a sufficiently detailed summary of the training content. Models presenting systemic risk carry evaluation, adversarial testing, incident reporting and cybersecurity duties in addition.

• Keep model documentation covering architecture, training and testing processes, and the evaluation results you rely on.
• Give downstream providers the information they need to meet their own obligations, in a form they can act on.
• Adopt and implement a policy for complying with Union copyright law, including the reservation of rights for text and data mining.
• Publish a sufficiently detailed summary of the content used for training, following the template of the AI Office.
• Systemic-risk models: run model evaluations and adversarial testing, track and report serious incidents, and protect the model and its physical infrastructure.

• AI Act Article 53: obligations of providers of general-purpose AI modelsArticle nodeOfficial text
• AI Act Article 55: obligations for models with systemic riskArticle nodeOfficial text
• AI Act Article 50: transparency for synthetic contentArticle nodeOfficial text

• How does the EU AI Act affect Indian generative AI and foundation models?
• Do general-purpose AI providers have to disclose their training data in the EU?

• AI Act: provider obligations
• Data transfers from the EU to India
• Indian software and SaaS handling EU data
• All requirements
• Indian law against EU law

Sector

Foundation models, LLM developers

Read date

2026-08-26

Register version

in-export-v1.0.0

Address

/in/export/gpai-models