OpenAI launches GPT-6 Cyber, a security model it says is stronger than GPT-6 Astra on cyber tasks and better aligned, first for Daybreak Red customers
Confirmed
Importance: major (4 of 5)Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- Major (4 of 5)
- Last verified
- 6 November 2026
Your AI and this story
- GPT-6 Astra189 days after its cutoff
- Claude Opus 5.5128 days after its cutoff
- Gemini 4 Argon219 days after its cutoff
- Grok 4.7158 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 128 days before it.
Key facts
- Availability: limited Daybreak Red customers first (CyberScoop)
- OpenAI claims: 81+ novel Firefox vulnerabilities; developed and validated a critical RCE in Brotli; ‘a major step up from GPT-5.6 Cyber’
- Safety: stricter identity and account security, more check-ins for high-risk requests, misalignment monitoring; trained with a simulated evaluation modelled on the Hugging Face sandbox escape (CyberScoop)
- Irregular (Nov 5): FrontierCyber 90/288 vs GPT-6 Astra 96 (neither solved the 15 hardest); CyScenarioBench 11/11 scenarios vs Astra 8/11, average success 61% (+7 pp); all 22 atomic challenges solved
- The Verge headline: ‘better at not hacking other companies’; TechCrunch: ‘Is that enough?’
- OpenAI’s X announcement: 2.86M views
What happened
OpenAI’s dedicated cyber model, rumoured before DevDay but not shown there, launched on Nov 5 to Daybreak Red red-teaming customers. openai.com returned 403 to us; figures come from CyberScoop and Irregular’s evaluation.
Why it matters
OpenAI pitches it as both more capable and more aligned, after a year of agent incidents (sandbox escapes, the Hugging Face intrusion). The independent numbers are mixed: better at multi-step attack scenarios, not better on open-ended challenges.
Sources
6 sources from 6 sites. Numbers match the chips in the text.
6 sources: 2 primary, 3 press, 1 reaction
Primary
- OpenAI: GPT-6 Cyber — advancing the frontier of aligned cybersecurity agentsopenai.com, official
- Irregular: assessing GPT-6 Cyber against offensive security benchmarksirregular.com, paper
Press
- CyberScoop: OpenAI’s GPT-6 Cybercyberscoop.com, press
- The Verge: OpenAI says its new cybersecurity model is better at not hacking other companiestheverge.com, press
- TechCrunch: OpenAI released a new cybersecurity model. Is that enough?techcrunch.com, press
Reactions
- OpenAI on Xx.com, discussion
Changes
- Filed