Cybersecurity — 2026-04-14
Treasury and Fed Convene Emergency Meeting with Bank CEOs Over Anthropic Mythos AI Cyber Risks
Treasury Secretary Bessent and Fed Chair Powell summoned Wall Street leaders to an emergency meeting on cyber risks posed by Anthropic's Mythos model, which the company delayed launching April 12 after determining it could lower barriers to catastrophic cybercrime against banking, healthcare, and energy infrastructure. Goldman Sachs, Citi, BofA, and Morgan Stanley are testing the model internally under Project Glasswing, a controlled distribution program for securing critical systems before similar AI capabilities proliferate. Anthropic limited initial access to roughly two dozen firms including JPMorgan, Amazon, and Apple.
Analysis
No single AI model release has ever prompted a joint Treasury-Fed emergency convening with bank CEOs; the meeting has no prior analog in financial regulation. The same administration that blacklisted Anthropic is directing its largest banks to test its most dangerous model, a logic that resolves only as a proliferation hedge: defenders need the capability before adversaries commoditize it. Project Glasswing may instead be congressional positioning rather than a genuine capability-threshold response.
No single AI model release has ever prompted a joint Treasury-Fed emergency convening with bank CEOs; the meeting has no prior analog in financial regulation. The same administration that blacklisted Anthropic is directing its largest banks to test its most dangerous model, a logic that resolves only as a proliferation hedge: defenders need the capability before adversaries commoditize it. Project Glasswing may instead be congressional positioning rather than a genuine capability-threshold response.
2 sources
- Wall Street Banks Try Out Anthropic Mythos -
Insurance Journal - Anthropic latest AI model Mythos triggers urgent risk review by UK regulators -
Business Today