Anthropic’s Mythos synthetic intelligence (AI) mannequin reportedly wanted solely hours to search out sure safety vulnerabilities in extremely delicate US authorities pc programs throughout an intelligence check, a US official informed the Related Press.
Nonetheless, that pace doesn’t imply it may exploit them in the identical window, the official stated, talking anonymously to debate the delicate matter.
Anthropic Mythos Mapped the Weak Spots in Safe Authorities Methods
The official stated the testing was carried out beneath Challenge Glasswing. Anthropic opted in opposition to a public launch for Mythos.
As an alternative, it granted a choose group of greater than 50 know-how corporations early entry to the unreleased mannequin, permitting them to establish and remediate vital software program vulnerabilities.
Senator Mark Warner of Virginia referenced the assessments on June 11. He spoke earlier than the Senate Committee on Banking, Housing, and City Affairs.
“This device broke into virtually all of our categorized programs, not in weeks however in hours,” he stated.
Warner attributed the account to the top of the Nationwide Safety Company (NSA) and US Cyber Command, Gen. Joshua Rudd.
Comply with us on X to get the most recent information because it occurs
Mythos already carries a observe document of discovering flaws. The UK’s AISI (AI Safety Institute) examined Mythos Preview on expert-level capture-the-flag challenges. The mannequin succeeded 73% of them. No mannequin had cleared that bar earlier than April 2025.
In April, Mozilla credited the AI mannequin with surfacing 271 vulnerabilities in Firefox. The browser maker patched them in Firefox 150.
Anthropic then launched Claude Fable 5 in early June. It billed the mannequin as a normal launch model of its Mythos tier, with added safeguards.
The opening was transient. On June 12, the US authorities issued an export management directive citing nationwide safety. The order required the agency to bar each international nationwide from Fable 5 and Mythos 5.
Subscribe to our YouTube channel to observe leaders and journalists present professional insights
The put up Anthropic Mythos Discovered Cracks within the Authorities’s Most Guarded Methods appeared first on BeInCrypto.