Skip to content
AI News HubLIVE
More
Summary only

GPT-5.5 Cyber Performance (as good as Mythos?)

AISI's evaluation of OpenAI's GPT-5.5 shows it matches or slightly exceeds Claude Mythos Preview on advanced cyber tasks, completing a 32-step corporate network attack simulation. It achieved a 71.4% pass rate on Expert-level tasks and solved a complex reverse-engineering challenge in 10 minutes. However, it failed an ICS attack simulation and a universal jailbreak was found by red team.

SourceHacker News AIAuthor: shmuli9
GPT-5.5 Cyber Performance (as good as Mythos?)
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

  • GPT-5.5 achieved 71.4% pass rate on Expert-level cyber tasks, compared to Mythos Preview's 68.6%.
  • It completed the 32-step 'The Last Ones' corporate network attack simulation in 2 of 10 attempts, second model to do so.
  • Failed the 7-step 'Cooling Tower' ICS attack simulation, but stalled on IT portion, not OT.
  • Red team discovered a universal jailbreak for all malicious cyber queries, later partially mitigated by OpenAI.

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

More recent stories on this topic · Models