US tech giants invited to discuss AI security tests at White House

The White House invited leading technology companies Meta and Anthropic to discuss a voluntary framework for cybersecurity tests on advanced AI models on Tuesday. This framework aims to give the government early access to the most sophisticated American AI models for testing hacking capabilities. The Trump administration finalized the details of these voluntary cybersecurity tests and plans to review them with the AI industry this week.

If the framework follows the details from the June executive order, companies that participate could provide their "covered frontier models" to the federal government for testing 30 days before allowing selected trusted partners to access them. The arrangement is intended to identify models with advanced cyber capabilities and determine how they can be safely deployed to enhance critical infrastructure.

However, the framework remains voluntary and does not grant Washington the authority to license models or block their public release. The development of the framework, directed by August 1, reflects the administration's attempt to introduce limited safeguards on increasingly capable AI systems while keeping American developers competitive with China, a concern that led Trump to abandon an earlier proposal for a longer review period.