Anime, manga, and games, with a take · A Yukimedia publication

← all stories otherannouncement 1 sources · 1h ago ·

OpenAI Splits Daybreak Into Blue and Red Tiers With GPT-5.6 Cyber

The split formalizes a dual-use approach where the unrestricted model is reserved for defensive researchers while the attack-tuned model is gated behind an application process.

Reporting from 1 source: GIGAZINE.

OpenAI Splits Daybreak Into Blue and Red Tiers With GPT-5.6 Cyber

OpenAI is expanding its Daybreak cybersecurity service with two new tiers. Daybreak Blue offers GPT-5.6 Sol without safeguards, while Daybreak Red provides GPT-5.6 Cyber, a model built on GPT-5.6 Sol with improved vulnerability discovery and exploit construction. OpenAI reports GPT-5.6 Cyber completed 95 percent of cyberattack instructions in testing and has already found real-world vulnerabilities.

OpenAI is dividing its Daybreak cybersecurity service into two tiers, each with its own model. Daybreak Blue gives access to GPT-5.6 Sol without safeguards, while Daybreak Red offers GPT-5.6 Cyber, which builds on GPT-5.6 Sol with added vulnerability discovery and exploit construction capabilities.

OpenAI's own tests show GPT-5.6 Cyber completed 95 percent of instructions to execute cyberattacks, compared with 1.5 percent for GPT-5.6 Sol and 2.0 percent for the unsafeguarded version. In exploit construction against the V8 JavaScript engine, GPT-5.6 Sol without safeguards scored highest when limited to 300 agent turns, but the gap narrowed at 600 turns.

OpenAI says GPT-5.6 Cyber has already found real-world flaws, including two V8 sandbox escapes, a local privilege escalation in a mobile OS, three remote code execution vulnerabilities in a widely used database, and more than 400 kernel vulnerabilities in a widely used OS. Access to both tiers requires an application through OpenAI's Trusted Access for Cyber form.

Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.

Sources