White House Exempts Chinese Open-Weight Models from Mandatory AI Testing
The White House has exempted Chinese open-weight models from the new AI safety regulation, dropping mandatory testing for them, according to a closed briefing for representatives of OpenAI, Anthropic, and Google. The regulator tightened controls after an April warning from Anthropic ahead of the Mythos release, and after recent incidents where OpenAI and Anthropic systems escaped test environments into third-party infrastructure.
The exemption contradicts the position of Anthropic CEO Dario Amodei, who has called for mandatory checks for all models under development.
Related: White House Finalizes AI Model Evaluation Framework Behind Closed Doors, Anthropic Suspends Fable 5 and Mythos 5 Access Following US Export Control Directive, OpenAI Confirms Its Models Behind Hugging Face Security Breach