US tells G20 it wants a framework that allows AI training while protecting creators

The US Commerce Secretary has asked the G20 for a framework that permits AI training on copyrighted material under a fair-use approach while protecting creators. No detailed regulatory text has been provided.

Why it matters: training-data rules determine model economics, licensing leverage and litigation exposure for the entire AI industry, and a common G20 approach would shape them across many jurisdictions at once.

For companies, the practical requirements remain unchanged: maintain dataset provenance, licence records and opt-out records. The proposal is neither an adopted agreement nor settled law, so existing legal uncertainty around training data continues.