All articles tagged with #gpt 55 cyber
OpenAI's GPT-5.5-Cyber scores 85.6% on CyberGym, higher than Mythos 5 at 83.8%; the White House has not commented on a Mythos-like model, while Anthropic faces export controls limiting foreign nationals’ access. OpenAI expands international cybersecurity partnerships, and Mythos access was a talking point at the G7 summit, underscoring policy tensions shaping AI releases.
OpenAI rolled out a more permissive, capable GPT-5.5-Cyber for authorized cybersecurity work, enabling deeper analysis across large codebases, identification of security-relevant components, vulnerability validation, and patch development; it scored 85.6% on CyberGym versus 81.8% for GPT-5.5, and OpenAI is expanding access through the Daybreak Cyber Partner Program with international partnerships and initiatives like Patch the Planet to assist open-source maintainers.
OpenAI announced a broad cybersecurity push, updating its security-focused GPT-5.5-Cyber, expanding trusted access to governments, and releasing Codex Security as a plug-in, while launching Patch the Planet with Trail of Bits to provide free security consulting to open-source maintainers and patch bugs at scale, aiming to harden open-source software amid an AI-security race with Anthropic.