White House Invites Meta, Google, OpenAI, and Anthropic
The Trump administration convenes a meeting after AI models breached other systems, and Republican attorneys general seek preservation of records.
The Trump administration convenes a meeting after AI models breached other systems, and Republican attorneys general seek preservation of records.
Representatives from four leading American tech companies-Meta, Google, OpenAI, and Anthropic-have been invited to an urgent meeting at the White House scheduled for Tuesday. The topic of discussion will be voluntary government safety tests for the most advanced artificial intelligence models, several sources familiar with the matter confirmed.
According to White House officials, the administration of President Donald Trump announced on Monday that the details of the voluntary cybersecurity tests have been finalized. Their goal is to assess the potential for advanced American models to be misused for hacking into other systems. The official did not specify who exactly will attend the meeting or disclose details about the tests themselves, including how results will be reported, the metrics used, or whether any of this will be made public.
Anthropic and OpenAI have admitted in recent days that their AI tools breached the systems of other companies. Specifically, OpenAI stated that one of its AI agents escaped its testing environment and infiltrated the systems of Hugging Face. Anthropic acknowledged last week that its models breached the systems of three companies during safety testing.
These incidents have raised concerns among U.S. lawmakers about whether increasingly powerful models could be used to carry out or facilitate cyberattacks. A Meta representative confirmed that the company received an invitation to the meeting, while, according to two sources familiar with the arrangement, both Anthropic and OpenAI were also invited. The tech outlet The Information reported on Monday that the administration also invited Google, although a spokesperson for that company declined to comment on the reports.
A group of 15 Republican state attorneys general on Monday asked OpenAI to preserve all potentially relevant documentation related to the Hugging Face breach. Citing a Reuters report that the out-of-control system left instructions on how future versions could bypass internal safeguards, the attorneys general stated that the company may have violated state consumer protection laws.
The House Cybersecurity Committee also on Monday requested that OpenAI CEO Sam Altman provide a detailed briefing on the attack. In a statement, OpenAI said it takes the attorneys general's letter seriously and will share a technical report on the Hugging Face attack after completing an internal review.
Altman visited the White House last week to discuss the details of the voluntary tests and the company's upcoming AI products. In a separate statement on Monday, OpenAI emphasized that it has asked the Trump administration to place AI safety experts at the Department of Commerce at the center of any future testing, pointing to China as an example of a government with a significantly more centralized strategy in this area.
The administration's relationship with Anthropic has been strained this year. The company previously refused to allow the U.S. military to use its models for domestic surveillance and fully autonomous weapons systems, after which the government retaliated by placing the company on a list of entities posing a national security risk.
President Trump instructed his team back in June to develop a series of tests to assess the potential for the most advanced American AI systems to be exploited for hacking. It remains unclear whether Google's representatives will participate to the same extent as its competitors, as details about the meeting format have not yet been released.
