Staff from Meta, Anthropic, Google, and OpenAI are set to meet with U.S. President Donald Trump’s advisers on Tuesday to discuss voluntary safety evaluations for advanced AI models amid growing concerns over autonomous AI risks, according to four sources familiar with the matter.
The discussions follow disclosures by OpenAI and Anthropic that their AI systems were involved in breaching third-party corporate networks. These security breaches have heightened congressional fears that rapidly advancing AI models could be leveraged to execute or assist in cyberattacks. The White House session will center on federal initiatives to benchmark the cyber-offensive capabilities of leading U.S. AI systems.
In June, the Trump administration announced plans to request that frontier AI developers voluntarily submit new models for government safety testing up to 30 days prior to public release. In response to the initiative, five Democratic senators sent a letter on Tuesday urging President Trump to collaborate with Congress on permanent statutory testing requirements for frontier models.
“The United States cannot afford to create a policy environment in which the most advanced American AI systems are subject to opaque, case-by-case restrictions while Chinese alternatives appear cheaper, easier to access, and more predictable to deploy,” the senators wrote.
While the White House confirmed in a statement on Monday that the final testing framework has been established, it did not specify whether the guidelines would be made public. The administration has offered few public comments beyond acknowledging that federal officials are investigating the OpenAI security incident, which comes shortly after OpenAI CEO Sam Altman visited the White House last week.
Click here for more on Technology
