Anthropic to share Mythos cyber flaw findings with global finance watchdog
2026-05-18 · AI (artificial intelligence) | The Guardian
Anthropic built an AI model so terrifyingly good at finding cyber vulnerabilities that they won't release it publicly and now have to go explain to global financial regulators just how screwed everyone's defenses might be. Nothing says 'everything is fine' like briefing the Financial Stability Board on your own creation's destructive potential. At least they're warning people before the hackers figure it out on their own.