OpenAI makes its rival to Anthropic's Mythos more widely available to cyber defenders
TL;DR
OpenAI is rolling out a more permissive version of GPT-5.5 — codenamed "Spud" — to vetted cyber defenders responsible for critical infrastructure. Recent testing shows GPT-5.5 nearly matches Anthropic's Mythos at finding and exploiting software bugs, fueling debate over misuse risks. Approved defenders in OpenAI's Trusted Access for Cyber program now get access to GPT-5.5-Cyber.
Nauti's Take
Promising: GPT-5.5-Cyber finally gives defenders a model on par with Anthropic's Mythos — ideal for red-team exercises and vulnerability hunting on critical infrastructure. The catch is dual-use risk: what defenders use legitimately today can leak or be misused tomorrow.
Teams in the Trusted Access program win big — everyone else should sharpen detection pipelines in parallel.
Sources