As AI-led attacks multiply, OpenAI launches a new cyber model
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI announced an expansion of Daybreak, its cyber defense service launched earlier this year, structuring it into two tiers—Blue and Red—both granting approved customers access to limited-access frontier cyber models.
- GPT-5.6 Cyber, a new model built off GPT-5.6 Sol, is available exclusively through the Red tier and is designed for specialized cybersecurity tasks including security testing and vulnerability research.
- GPT-5.6 Cyber access is currently limited to 'trusted customer partners' including Accenture, IBM, CrowdStrike, and Cloudflare, per the source.
- The Blue tier offers incident response, malware analysis, and patch validation, with OpenAI calling it the 'recommended starting point for most defenders.'
- The expansion comes shortly after Anthropic released its competing cyber-focused model Mythos, with the source noting AI labs are now competing to sell both the attack-capable models and the defenses against them.
- OpenAI had previously deployed significant guardrails limiting what customers could do with frontier models, and the Trump administration had previously sought to collaborate with AI companies on frontier model rollouts over safety concerns.
Why it matters: The 'trusted partner' gate (Accenture, IBM, CrowdStrike, Cloudflare) concentrates first access to GPT-5.6 Cyber among a handful of large vendors—a notable shift from OpenAI's earlier significant guardrails on what customers could do with frontier models, potentially widening the capability gap between large enterprise defenders and everyone else.
Ask SkimNews
