White House to host meeting on new AI review regime

The White House plans to host a meeting on Tuesday with AI executives to discuss a new voluntary system of government review for the technology's most powerful models, according to an administration official. The discussions will cover the framework outlined in President Donald Trump's executive order signed on June 2.

Under the central provision of this framework, companies like OpenAI, Google, and Anthropic would grant the government access to their most potent models for up to 30 days before releasing them to the public. This came in response to concerns over Anthropic's Mythos model, which the startup initially withheld from public release due to its potential to expose vulnerabilities in sensitive systems like those of banks, governments, and hospitals.

Mythos was eventually released with restrictions, but in June, the Commerce Department ordered Anthropic to suspend access to its Claude Fable 5 and Mythos 5 models for national security reasons linked to a reported jailbreak. Users were able to bypass guardrails designed to contain them, forcing the company to take both models offline for over two weeks before gradually restoring access.

This new AI assessment regime represents a shift in the Trump administration's stance on AI development. Initially, the administration had taken a hands-off approach, arguing that the government should not impede innovation in the global race with China to lead in AI.