Attack of the Killer Script Kiddies
Attack of the killer script kiddies
The Verge discusses Claude Mythos and AI bug finding, citing DARPA AIxCC scans over 54 million code lines. Teams found most seeded flaws plus over a dozen unseeded bugs; the RSS snippet does not disclose Mythos benchmarks, pricing, or access terms.
Why it matters: In DARPA's AI cyber challenge, Anthropic's Claude security test build Mythos scanned 54 million lines of code and found a dozen real bugs beyond the planted ones.