White House to test Anthropic's Mythos AI model
What's the story
The White House is preparing to distribute a version of Anthropic's new artificial intelligence (AI) model, Mythos, to major federal agencies. This was revealed in a memo, reviewed by Bloomberg, from Gregory Barbaccia, the federal chief information officer at the White House Office of Management and Budget. The Departments of Defense, Treasury, Commerce, Homeland Security, Justice and State have all been informed about this plan.
Deployment deliberations
Mythos protections being put in place
The memo also said that the Office of Management and Budget is working on putting protections in place for Mythos, so agencies can use it in the coming weeks. However, it's important to note that this plan is still under consideration and has not been finalized yet. More details are likely to emerge in the coming weeks.
Model concerns
Too dangerous for release
Last month, a leak suggested that Anthropic was developing a new AI model called Mythos. The company later confirmed the leak and warned that the model was so powerful it could pose a cybersecurity threat. Fearing potential misuse by hackers, Anthropic decided not to release it publicly for now. Instead, they launched Claude Opus 4.7, a less advanced version in terms of cyber capabilities than Mythos.
Limited access
Fear and shock over Mythos's capabilities
While the public got Opus 4.7, select companies like tech giants NVIDIA, Microsoft, Google and Apple as well as financial institutions such as JPMorganChase got limited access to a preview version of Mythos under an initiative called Project Glasswing. This has led to reports detailing the shock and fear among government and corporate officials who were allegedly allowed to view the model.
Information
Bank of England holding urgent discussions after previewing Mythos
The Bank of England is holding "urgent discussions" with cybersecurity officials after previewing Mythos, according to Financial Times. Meanwhile, The Information reported that the crypto industry has been scrambling to get its hands on this model to counter potential threats it could pose to cryptography.
Model applications
Briefings on offensive and defensive cyber applications
According to Bloomberg, Anthropic briefed senior officials across the US government before the limited release of Mythos. The briefings included possible offensive and defensive cyber applications of this powerful AI model. This comes as a stark contrast to just over a month ago when Anthropic's products were seen as a national security risk and effectively banned from federal use.