OpenAI Expands Daybreak Initiative with GPT-5.5-Cyber and Open-Source Patching
AI 안전 | Tue Jun 23 2026 00:00:00 GMT+0000 (Coordinated Universal Time) | 3 sources
OpenAI expanded its Daybreak initiative, announcing the general release of GPT-5.5-Cyber, a Codex Security update, and Patch the Planet, an open-source patching project.
Analysis
[OpenAI] announced expansion of the Daybreak initiative [1]
- Automating vulnerability discovery and patching
- Targeting major browsers
- network infrastructure
- FreeBSD
- and Linux kernel
- Providing frontier AI security capabilities to defenders
[GPT-5.5-Cyber] released the general version of its cybersecurity-specialized model [1][3]
- Achieved 85.6% on CyberGym benchmark
- SOTA performance compared to GPT-5.5's 81.8%
- Limited release to trusted defenders
[Codex Security] updated security plugin and distributed it as an app plugin [1][3]
- Accelerates vulnerability discovery and patching in existing systems
- Automatically blocks new vulnerabilities before they reach production
- Subsidized 20 trillion tokens for open-source and private code
[Patch the Planet] launched initiative to support open-source maintainers [1][2][3]
- Collaboration with Trail of Bits
- HackerOne
- and Calif
- Over 30 projects participating
- including cURL
- Go
- Python
- Sigstore
- and pyca/cryptography
- Provides free consulting from vulnerability validation through patch development and disclosure
[Trail of Bits] conducted security work on 19 projects via a 5-day open sprint [2][3]
- Deployed 25 engineers
- about 1/5 of total staff
- Identified hundreds of security issues and merged dozens of patches
- Built reusable infrastructure including fuzzing harnesses and CVE analysis pipelines
[Daybreak Cyber Partner Program] provided trusted access to frontier models for security partners [1][2]
- Integrates OpenAI's top-performing models into partner products and services
- Access based on governance and human oversight
- Provides ChatGPT Pro and API credits to participating open-source projects