technologyScore 60/100Research
Amodei: Anthropic's 'Mythos' model given early to ~40 firms incl. Apple, Microsoft, Google, banks to patch bugs before public release
“We've given the model to, I think it's something like 40 companies now... one of the well-known companies found more bugs and fixed more bugs with Mythos in a week than they had for the last year.”
Why it matters
A frontier AI model is now superhuman at exploit discovery, and its release is being gated on pre-patching critical infrastructure at major banks and tech firms — a new, measurable step-function in AI-driven cybersecurity economics.
Investment implication
If AI vulnerability discovery becomes standard, cybersecurity vendors that productize AI-driven code auditing and patching (e.g., CrowdStrike, Palo Alto, Checkmarx-style SAST/SCA players) could see demand accelerate; conversely legacy pen-testing services face disruption. Anthropic itself is private with no direct equity proxy.