Logan Graham
1.3K posts
@logangraham
Head of the Frontier Red Team @anthropicai. 🌎 Make things radically good.
The security landscape is changing fast. I'm hiring people to figure out what comes next. My team led this work — and we're expanding.
We partnered with Mozilla to test Claude's ability to find security vulnerabilities in Firefox. Opus 4.6 found 22 vulnerabilities in just two weeks. Of these, 14 were high-severity, representing a fifth of all high-severity bugs Mozilla remediated in 2025.
We partnered with Mozilla to test Claude's ability to find security vulnerabilities in Firefox. Opus 4.6 found 22 vulnerabilities in just two weeks. Of these, 14 were high-severity, representing a fifth of all high-severity bugs Mozilla remediated in 2025.
We didn't do anything too fancy. Probably the key thing for us was using task verifiers to let the model verify whether it had achieved the goal (e.g. a crash, or correctly builds post-patch) Just believe in the power of the models, give them tools to use, and walk away!
Now is a good time to say I'm hiring @anthropicai for the Frontier Red Team. We need Research Scientists on the biggest issues in model safety, like cyber, autonomy, and agent risks. 2026 is the year. I can promise you your life's work and the most meaningful mission.
