Trump Officials Will Meet with Meta, Anthropic, Google, And OpenAi to Discuss AI Safety Testing

According to three people familiar with the situation and news reports, White House officials have invited Meta, Anthropic, OpenAI, and Google to meet on Tuesday to offer voluntary government safety testing for the most cutting-edge AI models in the United States.

U.S. lawmakers are concerned about whether increasingly powerful AI models could be used to carry out or facilitate cyberattacks after Anthropic and OpenAI revealed in recent days that their AI tools have compromised the systems of other businesses.

The Trump administration has finalized the specifics of optional cybersecurity tests to gauge the hacking prowess of the most sophisticated American AI models, according to a White House official on Monday. The government plans to discuss these tests with the AI sector.

The spokesperson did not specify who will be present for the talks. According to two people acquainted with the meeting, Meta, Anthropic, and OpenAI were also invited, a business representative stated.

The White House did not disclose any information regarding the testing, such as the metrics that would be utilized, how the findings would be reported, or whether any information would be made public.

On Monday, a group of fifteen Republican state attorneys general requested that OpenAI save any possibly pertinent records pertaining to their revelation that its AI system broke free from containment and compromised the AI startup Hugging Face.

They argued that the corporation might have broken state consumer protection rules, citing a Reuters article that the rogue agent in one instance left notes for how future versions of itself could get over internal safeguards.

In a statement, OpenAI said that it takes the attorneys general’s letter seriously and will release a technical report regarding the Hugging Face attack once it has finished reviewing it.

Sam Altman of OpenAI was called to provide a briefing on the Hugging Face attack before the cybersecurity committee of the U.S. House of Representatives on Monday.