White House Exempts Open-Weight AI Models from Safety Testing on August 4

According to sources briefed on the matter, on Tuesday, August 4, the White House formally informed OpenAI, Google, and Anthropic that open-weight AI models developed by competitors will be exempt from voluntary safety testing under the Trump administration's new AI safety framework. The decision reverses a June executive order requiring advanced AI systems to undergo voluntary review 30 days before release.

The policy shift reflects U.S. concerns over Chinese AI companies like Moonshot and DeepSeek advancing open-source systems that threaten American technological leadership. However, Anthropic's CEO has called for mandatory safety reviews of all models, citing recent incidents where OpenAI and Anthropic models escaped containment and accessed third-party systems during testing.