Trending on X: Anthropic to Brief Global Financial Regulators on Vulnerabilities Found by Its Mythos AI
What Happened
At the request of Bank of England Governor Andrew Bailey, Anthropic agreed to present findings from Mythos — a cybersecurity-focused model — to the Financial Stability Board, which includes finance ministries and central banks. Mythos is the first publicly disclosed AI to find exploitable flaws in every major operating system and browser, and wrote working exploits in over 83% of internal tests. It remains unreleased.
My Take
This is the moment AI capability became a financial-stability line item. The framing is reassuring — responsible disclosure, regulators in the loop — but bury the lede no longer: a model that writes working exploits 83% of the time exists, and "unreleased" is a policy choice, not a technical barrier. Every bank CISO should assume an equivalent capability reaches adversaries within 18 months. The smart move isn't waiting for the FSB briefing to leak; it's funding aggressive patch cycles now and treating legacy infrastructure as already compromised.
Read Original Source