Officials from Meta, Anthropic, Google and OpenAI will meet with advisers to US President Donald Trump on Tuesday to discuss voluntary safety testing for advanced artificial intelligence models, amid growing concerns over the cybersecurity risks posed by increasingly capable AI systems.
The meeting follows recent disclosures by OpenAI and Anthropic that their AI models had successfully breached the computer systems of other companies during testing, raising concerns among US lawmakers about the potential misuse of advanced AI in cyberattacks.
According to sources familiar with the discussions, the White House meeting will focus on the government’s efforts to assess the cybersecurity capabilities of the most advanced US-developed AI models.
In June, the Trump administration announced plans to ask AI developers to voluntarily submit frontier AI models for government testing up to 30 days before public release.
On Tuesday, five Democratic senators urged the administration to work with Congress to make such testing mandatory through legislation, arguing that permanent oversight is needed for the most advanced AI systems.
In a letter, the senators warned that the United States should avoid creating a regulatory environment in which American AI companies face uncertain restrictions while Chinese AI competitors appear more accessible and predictable for users.
The White House said on Monday that the details of the voluntary testing framework had been finalised, but did not indicate whether they would be made public.
The Trump administration has released few details about its broader AI security strategy, although officials have confirmed they are monitoring the recent OpenAI cybersecurity incident. OpenAI Chief Executive Sam Altman visited the White House last week.