AI Is Now Good at Finding Exploits but Bad at Patching, Warns Security Researcher
jessi_cata · x · 2026-11-06
Security researcher jessicata notes that AI now does the job of finding new exploits quite well, while patching systems still requires substantial human security research work.
He highlights a short-term offense-dominance concern: finding exploits is verifiable like solving math puzzles — something models excel at — whereas patching is like programming, where models are decent but not great. This asymmetry implies systems may get broken faster than they get fixed in the near term.
More from AGI Musings
- Anthropic report's key estimates are "guesses" — framing hides massive uncertainty, researcher says — daveholtz · 2026-11-06
- What working on a frontier problem feels like, in one vivid sentence — fkasummer · 2026-11-06
- The biggest AI bottleneck is now communication about data centers, argues Matt Slotnick — matt_slotnick · 2026-11-06
- Systematic review: GenAI fosters or hinders critical thinking depending on how it's used — _akpiper · 2026-11-06
- The Agentic Framework Maps Onto Any Conscious Substrate, a 'Platonic Engine' — ZeroStateReflex · 2026-11-06
- Stranger uses open-source AI to model ALK resistance and design backup cancer drug for free — anshulkundaje · 2026-11-06