The White House has opposed Anthropic's plans to expand the use of its artificial intelligence (AI) model Mythos, complicating the rollout of an AI tool capable of launching cyberattacks and causing widespread network chaos. Anthropic recently proposed adding about 70 more companies and organizations to use Mythos, bringing the total number of entities with access to about 120, people familiar with the matter said.
U.S. government officials told the company they opposed the move because of security concerns, these people said. Some White House officials are also concerned that Anthropic doesn’t have enough computing power to serve so many new entities without affecting the government’s ability to use the model effectively, one of the people said.
Because the Mythos model posed a national security risk, the White House became involved in its promotion. The model's ability to find and exploit software vulnerabilities has alarmed government agencies and companies in recent weeks. Discussions between Anthropic and the White House are aimed at addressing these risks and trying to repair the relationship between Anthropic and the government.
The White House's rejection of Anthropic's request to let more organizations use Mythos shows that despite efforts by both sides to de-escalate the situation, the relationship remains complicated. The U.S. government has tried to sever ties with Anthropic as the company feuds with the Pentagon over how the military uses its AI tools, a battle that is working its way through the legal system through two separate court cases.
Anthropic is initially open for use to about 50 companies and organizations that manage critical infrastructure, and there are no plans for a public release. Some government agencies have access, and the Trump administration is working to expand that.
A White House official said the U.S. government is balancing innovation and safety when working with private industry to ensure that AI models are promoted safely.
