US agency pursues access to new Anthropic AI model

The US Treasury Department is reportedly seeking access to Anthropic’s restricted Mythos model amid growing concern over advanced AI dramatically reshaping the cybersecurity threat landscape. Bloomberg reported US Treasury Secretary Scott Bessent and Federal Reserve chair Jerome Powell briefed Wall Street executives earlier this month on the potential negative impact Mythos could have across the cybersecurity sector. The US Treasury Department’s technology team is apparently seeking access to the Mythos AI model to search for vulnerabilities in its software and computer systems. In March, the Department of War (DoW) designated Anthropic as a supply-chain risk after the AI company refused to grant the  Pentagon unfettered rights  to deploy its models over concerns about domestic surveillance and use in lethal autonomous weapons, Project Glasswing The AI startup introduced its Claude Mythos model on 7 April, under the auspices of its Project Glasswing, to a limited number of technology companies including Amazon Web Services, Apple, Nvidia and Google instead of making it publicly available. Anthropic stated Mythos Preview is capable of autonomously identifying and exploiting software vulnerabilities at a level which surpasses most human experts. Participants in Project Glasswing will use the model defensively to strengthen their systems, with findings shared more broadly. During testing, it reportedly uncovered thousands of previously unknown “zero‑day” flaws across major operating systems and web browsers, vulnerabilities which attackers could exploit before developers can issue patches. The company warned Mythos could enable more frequent and damaging cyberattacks against companies and critical infrastructure. While Anthropic believes AI will ultimately favour defenders by hardening software at scale, it noted in a blog the transition period will be volatile, as hackers and nation states rapidly adopt similar capabilities while patching efforts lag discoveries. The post US agency pursues access to new Anthropic AI model appeared first on Mobile World Live .