Overview
- Treasury Secretary Scott Bessent and Fed Chair Jerome Powell met bank chiefs at Treasury on Tuesday to flag risks from Anthropic’s new Mythos model, with CEOs from Citigroup, Morgan Stanley, Bank of America, Wells Fargo, and Goldman Sachs present and JPMorgan’s Jamie Dimon unable to attend.
- Anthropic says Mythos can spot and weaponize software flaws at scale, including a 27-year bug in OpenBSD and a weakness in FFmpeg that automated tools scanned five million times without catching.
- Access to Mythos is tightly restricted under Project Glasswing for defensive use by a vetted group that includes Apple, Google, Microsoft, Nvidia, Amazon, JPMorgan, and key open‑source stewards, with up to $100 million in credits and $4 million for security groups to fix vulnerabilities before attackers do.
- In a parallel legal fight, a Washington, D.C., appeals court on Wednesday refused to pause the Pentagon’s “supply‑chain risk” label on Anthropic but ordered a fast‑track schedule, diverging from a San Francisco judge’s earlier injunction that blocked a separate order tied to Defense contracts.
- Regulators now treat advanced AI that automates hacking as a potential systemic threat to core banking systems, a shift that could drive rapid upgrades to cybersecurity tools and incident response across the largest U.S. lenders.