Source: Xinhua

MOSCOW, July 8 — Russia's State Duma passed a law to establish a legal framework for the development, deployment and use of large-scale artificial intelligence foundation models. The legislation creates the categories of "sovereign" and "national" AI foundation models and grants developers access to state data for training purposes.

A sovereign AI model must be developed and fully controlled throughout its life cycle by a Russian legal entity and hosted in Russian data centers, while a national AI model must also be developed by a Russian legal entity but may incorporate foreign open-source components. Both categories are required to comply with Russian laws and the country's traditional moral and spiritual values. The government is authorized to require exclusive use of sovereign and national models in government information systems and other sensitive sectors. The law requires audio and visual content generated by large AI foundation models to be labeled. Most provisions take effect on Sept. 1, whereas requirements concerning sovereign and national models, developer obligations, labeling of AI-generated content and the use of intellectual property in model training enter into force on March 1, 2027. Existing systems will be exempt until Sept. 1, 2032 provided data are processed and stored within Russia.