White House opposes Anthropic's expansion of Mythos usage to 120 companies, citing concerns over insufficient computing power

Odaily Planet Daily News The White House recently opposed Anthropic's proposal to expand its AI model Mythos to approximately 120 companies, primarily based on security and computing power considerations. Anthropic originally planned to add 70 new companies to use Mythos on top of the existing approximately 50 companies, but the White House raised questions about this, fearing that insufficient computing power could affect the government's own use of Mythos.

Released in early April, Mythos is designed to detect and exploit critical software vulnerabilities and is currently limited to enterprise testing for managing critical infrastructure, with no public release plan in place. The White House is concerned that expanding usage to more commercial users will lead to computing power bottlenecks for the government when using the model, particularly regarding Anthropic's computing power procurement agreements with Amazon, Google, and Broadcom. Although agreements have been signed, new capacity has not yet come online.

On the political front, relations between the White House and Anthropic have not eased. The Trump administration has publicly criticized Anthropic for hiring multiple former Biden administration officials and expressed dissatisfaction with its ties to liberal organizations. One example demonstrates the trust issues between the two sides: Collin Burns, a former researcher at Anthropic, was originally assigned to a government AI model evaluation position, but senior White House officials directly replaced the candidate upon learning of it, citing the need to avoid AI company personnel directly engaging in matters involving dealings with other AI companies.

Additionally, Anthropic last week disclosed unauthorized access incidents involving the Mythos model, further intensifying external regulatory scrutiny of the company.