Skip to content
AI News HubLIVE
More
Summary only

ExploitGym: Can AI agents turn bugs into exploits?

ExploitGym is a large-scale, diverse, and realistic benchmark for evaluating AI agents' ability to turn security vulnerabilities into working exploits. It comprises 898 instances from real-world vulnerabilities across userspace programs, Google's V8 JavaScript engine, and the Linux kernel. Evaluation shows that frontier models like Claude Mythos Preview (157 successes) and GPT-5.5 (120) can exploit a non-trivial fraction of vulnerabilities, even with common defenses enabled, highlighting growing cybersecurity risks.

SourceHacker News AIAuthor: p_stuart82
ExploitGym: Can AI agents turn bugs into exploits?
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions

Summary only

Under the source content permission, this page provides an editorial summary, key points, and a link to the original article.

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • ExploitGym benchmarks AI agents on converting vulnerabilities to exploits.
  • Includes 898 real-world instances from userspace, V8, and Linux kernel.
  • Frontier models Claude Mythos Preview and GPT-5.5 achieved 157 and 120 successful exploits.
  • Non-trivial success persists even with commonly used defenses.

Highlights and analysis are generated automatically and may contain errors. Check the original source.

More recent stories on this topic · Models