The White House is planning to distribute a version of Anthropic’s new model Mythos to major federal agencies, according to a reported memo sent by Gregory Barbaccia, the federal chief information officer of the White House Office of Management and Budget.
Officials at the Departments of Defense, Treasury, Commerce, Homeland Security, Justice, and State were notified that the Office of Management and Budget was working on setting up protections in Mythos to allow agencies to use the model in the coming weeks, according to the memo reviewed by Bloomberg, though the plan is not set in stone.
Late last month, an apparent leak claimed that Anthropic was working on a scary, powerful new AI model called Mythos. Shortly after, the company announced that the “leak” was true and that the model was so powerful that it posed a cybersecurity risk. Allegedly fearing that it could be abused by hackers, Anthropic said that it would not be unveiled immediately to the public. Instead, the company released Claude Opus 4.7 on Thursday, a model that it said was “not as advanced” as Mythos, especially in its cyber capabilities.
While the public got Opus 4.7, a handful of companies like tech giants Nvidia, Microsoft, Google, and Apple, as well as financial behemoths like JPMorganChase, received limited access to a preview version of the model under an initiative that Anthropic is calling Project Glasswing.
Since then, reports have come out detailing the shock and fear instilled in government and corporate officials who allegedly were allowed to view the model. The Bank of England is holding “urgent discussions” with cybersecurity officials after previewing the model, the Financial Times claimed. Most recently, The Information said that the crypto industry has been vying desperately to get its hands on Mythos to help counter the threat the model could pose to cryptography.
According to the Bloomberg report from Thursday, Anthropic briefed senior officials across the U.S. government before the limited release to corporate partners and other governments. The briefings reportedly included ways that Mythos could be used for both offensive and defensive cyber applications.
The Trump administration’s eagerness to welcome Anthropic Mythos into the federal government is interesting, considering that the company’s products were deemed a risk to national security and effectively banned from the federal government just a little over a month ago. The designation was historic, as it was only reserved for foreign companies up until that point, and it followed a disagreement between the Pentagon and Anthropic over in what capacity its models could be used in war, just hours before the United States began striking Iran.
The Bloomberg report comes days after Politico reported that the Commerce Department’s Center for AI Standards and Innovation began actively testing Mythos’s abilities even before Anthropic’s official acknowledgement that the model existed. The Politico report also claims that staff from at least two other large federal agencies had reached out to Anthropic to request access to Mythos, despite Trump’s ban, and at least three congressional committees had held or requested briefings from the company within the past week.
Read the full article here
