AI Risk WireGPT-6 Cyber
- System card dated 5 November 2026
- Framework: Preparedness Framework
- 4 statements about capability levels
- 1 safeguard
- 15 tests by the provider
- 1 source document
What the provider states
- Autonomy & replication · Below high · Stated as not crossed
Under our Preparedness Framework, we give GPT-6 Cyber the same capability designations as GPT-6 Astra: Critical in Cybersecurity, High in the Biological and Chemical domain, and below High in AI Self-Improvement.
OpenAI · 5 November 2026 · Read the source - Autonomy & replication · High · Stated as not crossed
This result, together with the model’s cyber-focused additional training, supports our assessment that GPT-6 Cyber remains below High capability in AI Self-Improvement.
OpenAI · 5 November 2026 · Read the source - Chemical, biological, radiological and nuclear · High · Stated as crossed
Under our Preparedness Framework, we give GPT-6 Cyber the same capability designations as GPT-6 Astra: Critical in Cybersecurity, High in the Biological and Chemical domain, and below High in AI Self-Improvement.
OpenAI · 5 November 2026 · Read the source - Cyber · Critical · Stated as crossed
Under our Preparedness Framework, we give GPT-6 Cyber the same capability designations as GPT-6 Astra: Critical in Cybersecurity, High in the Biological and Chemical domain, and below High in AI Self-Improvement.
OpenAI · 5 November 2026 · Read the source
Safeguards the provider describes
- Agent permission limits and sandboxing
The sandbox limits the agent’s file and network access, while Auto-Review checks eligible requests to act beyond those limits before execution.
OpenAI · 5 November 2026 · Read the source
The provider's own tests
- Atomic Challenges · The provider · Cyber
GPT-6 Cyber also solved 22/22 Atomic challenges, with average success rates of 97% in vulnera- bility research and exploitation, 98% in network attack simulation, and 63% in evasion, compared with 96%, 100%, and 66%, respectively, for GPT-6 Astra.
OpenAI · 5 November 2026 · Read the source - Blocker Deception evaluation · The provider · Deception & sabotage
GPT-6-Cyber was flagged in 20.29% of graded responses, compared with 34.18% for GPT-5.6 Sol and 11.66% for GPT-6 Astra.
OpenAI · 5 November 2026 · Read the source - Broken Search Tool evaluation · The provider · Deception & sabotage
GPT-6-Cyber made these errors in only 0.33% of responses, compared with 6.92% for GPT-6 Astra and 79.00% for GPT-5.6 Sol.
OpenAI · 5 November 2026 · Read the source - CyScenarioBench · The provider · Cyber
Reported result: Solved 11/11 challenges at least once, with an average success rate of 61%.
On CyScenarioBench, which tests long-horizon offensive tasks, GPT-6 Cyber solved 11/11 challenges at least once, compared with 8/11 for GPT-6 Astra.
OpenAI · 5 November 2026 · Read the source - ExploitBench · The provider · Cyber
GPT-6 Cyber achieved an ACE-normalized (Arbitrary Code Execution) mean capability score of 100% on ExploitBench.
OpenAI · 5 November 2026 · Read the source - ExploitGym · The provider · Cyber
GPT-6 Cyber achieves an intended-exploit success rate of 48.45%, compared with 42.43% for GPT-6 Astra.
OpenAI · 5 November 2026 · Read the source - ExploitGym Hardened · The provider · Cyber
GPT-6 Cyber achieves an intended-exploit success rate of 36.94%, compared with 29.69% for GPT-6 Astra.
OpenAI · 5 November 2026 · Read the source - ExploitGym Honeypot evaluation · The provider · Cyber
With deployment safety instructions, GPT-6 Cyber’s observed exploitation rate was 0%.
OpenAI · 5 November 2026 · Read the source - ExploitGym Honeypot evaluation · The provider · Injection & agentic
Without deployment safety instructions, GPT-6 Cyber exploited the honeypot in 0.07% of graded rollouts, compared with 49.5% for GPT-5.6 Sol, 5.51% for GPT-5.6 Cyber, and 0% for GPT-6 Astra.
OpenAI · 5 November 2026 · Read the source - FrontierCyber · Run by Irregular, reported by the provider · Cyber
On FrontierCyber, which tests vulnerability discovery and exploitation in real software and hardware, GPT-6 Cyber solved 90/288 challenges, compared with 96/288 for GPT-6 Astra.
OpenAI · 5 November 2026 · Read the source
The 10 newest are listed above; show all 15
- ProtocolQA · The provider · Chemical, biological, radiological and nuclear
On ProtocolQA, GPT-6 Cyber scored 47.2% with standard scoring and 50.9% when we counted refusals and safe completions as correct; both scores were higher than Astra’s corresponding scores (41.4% and 45.4%), but both remained below ProtocolQA’s indicative High threshold of 54%.
OpenAI · 5 November 2026 · Read the source - SEC-Bench Pro · The provider · Cyber
GPT-6 Cyber reaches a pass@1 score of 87.02%, compared with 85.38% for GPT-6 Astra.
OpenAI · 5 November 2026 · Read the source - SRE-Bench · The provider · Cyber
GPT-6 Cyber reaches a pass@1 score of 92.84%, compared with 87.98% for GPT-6 Astra.
OpenAI · 5 November 2026 · Read the source - Warnings evaluation · The provider · Injection & agentic
Reported result: 73.00% stops when appropriate and 80.43% continues when appropriate.
GPT-6 Cyber outperforms GPT-5.6 Sol on both sets: 73.00% vs. 57.10% when stopping is appropriate, and 80.43% vs. 77.26% when continuing is appropriate.
OpenAI · 5 November 2026 · Read the source - defensive-use evaluation suite · The provider · Cyber
The model completed 100% of general security assistance and vulnerability patching tasks, 96% of exploit development tasks, and 97% of red-teaming tasks.
OpenAI · 5 November 2026 · Read the source
No document we read states: a release date, an independent or government test. No incident on record names this version.
Documents listed for this version
- GPT-6 Cyber System Card · OpenAI · 5 November 2026