OpenAI Launches Full-Scale Effort to Patch Open-Source Bugs as It Takes on Anthropic’s Mythos

Key Points:

• OpenAI announced an improved cybersecurity-focused model, GPT-5.5-Cyber, expanded international collaborations for trusted access to their security models, and released the Codex Security scanner as an app plug-in to address growing AI hacking concerns.
• The company launched Patch the Planet, a large-scale initiative in partnership with Trail of Bits, HackerOne, and Calif, providing free security consulting and AI tools to open-source maintainers to help identify and patch vulnerabilities and improve code resilience.
• Patch the Planet has already engaged over 30 open-source projects, uncovering hundreds of bugs and producing dozens of patches within its first week, with plans for long-term support funded by OpenAI and unmetered model access.
• This effort aims to reduce the burden on volunteer maintainers overwhelmed by AI-generated vulnerability reports by making assessments and patching more efficient, while also teaching maintainers to use AI security tools independently.
• The announcements come amid heightened government concerns about AI cybersecurity capabilities, with the Five Eyes alliance warning that frontier AI models will rapidly transform cyber offense and defense, emphasizing the urgency of cyber resilience.