NewsBite analysis · AI, Security
The Decoder
OpenAI says new GPT-5.5-Cyber outperforms Anthropic's Mythos on cybersecurity benchmark
OpenAI has introduced GPT-5.5-Cyber, designed to enhance cybersecurity with improvements to its Codex Security plugin. This initiative aims to bridge the gap between identifying vulnerabilities and effectively patching them. The updated model has reportedly set new records in cybersecurity benchmarks. With a partner network of over 25 security firms and expanded government collaborations, the initiative is geared to provide robust security solutions across various platforms.
Act nowConfidence: HighShort term horizon
Published 23 Jun 2026, 12:10 UTC3 min read
Visit original article ↗Lead image — The Decoder
Summary
TL;DR
- GPT-5.5-Cyber outperforms Anthropic's Mythos in security benchmarks.
- OpenAI's expanded partner network includes 25+ security firms.
- Codex Security plugin automates vulnerability discovery and patching.
- Open source initiative 'Patch the Planet' launched.
- Model designed to assist developers in identifying and fixing vulnerabilities.
Interesting facts
- GPT-5.5-Cyber includes deep scans of entire codebases and attack path analysis.
- The plugin has scanned over 30 million commits across 30,000 codebases.
- OpenAI has Trusted Access partnerships with multiple countries.