The White House is set to host a meeting on Tuesday with leaders from major AI companies to discuss a new U.S. framework for voluntary safety and cybersecurity testing of the most powerful AI models, according to multiple reports. Outlets say the meeting includes representatives from OpenAI, Anthropic and Google, with other companies also mentioned, such as Meta, alongside technology firms like Nvidia and Microsoft.
The framework stems from an executive order signed by President Donald Trump on June 2. Under the central provision described by reports, companies provide the government access to their advanced models for up to 30 days before the models are released to the public. Several outlets also report that details of the framework have not been released publicly, and the administration is described as keeping aspects of the process or benchmarks confidential.
The change in approach is linked to national security and cybersecurity concerns raised during the administration’s earlier, more hands-off posture on AI development. One trigger cited across coverage involves Anthropic’s Mythos model, its later restriction on foreign access tied to a reported jailbreak, and the subsequent suspension and restoration of access to related models after regulatory action by the Commerce Department.
