AnswersPublished19 sources
What did OpenAI announce about GPT-6 Cyber on Thursday, November 5, including its status as the company’s first cybersecurity model rated “Critical,” what that highest Preparedness Framework risk tier means for zero-day exploitation and end-to-end attacks on real-world critical s
OpenAI announced GPT-6 Cyber on Thursday, November 5, 2026, releasing a model designed to find and exploit security flaws to a small group of vetted customers and rating it “Critical” for cybersecurity under its Preparedness Framework. [2] The available excerpts verify parts of the announcement, but OpenAI announced...
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- OpenAI announced GPT-6 Cyber on Thursday, November 5, 2026, releasing a model designed to find and exploit security flaws to a small group of vetted customers and rating it “Critical” for cybersecurity under its Preparedness Framework. [2] The available excerpts verify parts of the announcement, but
- OpenAI announced GPT-6 Cyber on Thursday, November 5, 2026, releasing a model designed to find and exploit security flaws to a small group of vetted customers and rating it “Critical” for cybersecurity under its Preparedness Framework. [2] The available excerpts verify parts of t
- ## Rating and intended use
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What did OpenAI announce about GPT 6 Cyber on Thursday, November 5, including its status as the company’s first cybersecurity model rated “C. Article summary: OpenAI announced GPT 6 Cyber on Thursday, November 5, 2026, releasing a model designed to find and exploit security flaws to a small group of vetted customers and rating it “Critical” for cybersecurity under its Prepared. Topic tags: general web, ai safety, openai, agents, ai. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts w
OpenAI announced GPT-6 Cyber on Thursday, November 5, 2026, releasing a model designed to find and exploit security flaws to a small group of vetted customers and rating it “Critical” for cybersecurity under its Preparedness Framework. 2 The available excerpts verify parts of the announcement, but provide insufficient evidence to independently confirm every detail in your question.
Rating and intended use
- GPT-6 Cyber is a cybersecurity-specialized variant of GPT-6 Astra, intended for authorized vulnerability discovery, exploit development, and web penetration testing.
6
- It is reported as OpenAI’s first cybersecurity-specific model rated “Critical,” compared with GPT-5.6 Cyber’s “High” rating; that does not mean it is OpenAI’s first model overall to reach “Critical.”
8 OpenAI’s system-card excerpt says GPT-6 Cyber shares GPT-6 Astra’s “Critical” cybersecurity designation.
5
- Insufficient evidence: the excerpts do not define the highest tier’s precise thresholds for zero-day exploitation or end-to-end attacks on real-world critical systems, nor explain the full distinction from “High.”
Vulnerability research and evaluations
- OpenAI’s announcement excerpt says researchers used GPT-6 Cyber to develop 948 Linux privilege-escalation exploits and identify 81 previously unknown high-severity Firefox vulnerabilities.
9
- Reporting also identifies 588 of the Linux vulnerabilities as previously unknown.
13
- GPT-6 Cyber scored 48.5% in OpenAI’s ExploitGym evaluation.
- OpenAI reports that the model avoided exploiting out-of-scope services in 99.93% of alignment tests without its deployment prompt and 100% with that prompt—a test result, not a guarantee of perfect behavior in deployment.
- Insufficient evidence: the supplied excerpts do not verify Mozilla’s patching progress, the precise local-root implications of all 588 vulnerabilities, the more-than-fivefold comparison with July’s GPT-5.6 Sol research, the confirmed Brotli remote-code-execution flaw, GPT-5.6 Cyber’s 32.1% score, or OpenAI’s reported refusal to describe the 0.07% of scope violations.
Access, safeguards, and broader context
- The initial release was restricted to vetted customers.
12
- OpenAI’s developer documentation describes using the model to investigate vulnerabilities, reproduce them in authorized test environments, and verify fixes.
1
- Insufficient evidence: the excerpts do not establish simultaneous access to GPT-5.6 Cyber for all Daybreak Red customers, the exact expansion timetable, stricter identity and account-security rules, high-risk-task check-ins, monitoring requirements, or training specifically informed by sandbox escapes and the hacking of Hugging Face.
- Insufficient evidence: they also do not verify the “fourth cybersecurity-focused model this year” count, competing Anthropic and Google vetted-access programs, Gartner’s September $51.3 billion forecast, or the complete stated rationale concerning increasingly malicious AI use.
The supported picture is a restricted-access model for authorized security research, with stronger reported exploit capabilities and high measured adherence to task boundaries; the remaining requested details cannot be confirmed from the excerpts provided. 6
9
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "What did OpenAI announce about GPT-6 Cyber on Thursday, November 5, including its status as the company’s first cybersecurity model rated “Critical,” what that highest Preparedness Framework risk tier means for zero-day exploitation and end-to-end attacks on real-world critical s"?
OpenAI announced GPT-6 Cyber on Thursday, November 5, 2026, releasing a model designed to find and exploit security flaws to a small group of vetted customers and rating it “Critical” for cybersecurity under its Preparedness Framework. [2] The available excerpts verify parts of the announcement, but
What are the key points to validate first?
OpenAI announced GPT-6 Cyber on Thursday, November 5, 2026, releasing a model designed to find and exploit security flaws to a small group of vetted customers and rating it “Critical” for cybersecurity under its Preparedness Framework. [2] The available excerpts verify parts of the announcement, but OpenAI announced GPT-6 Cyber on Thursday, November 5, 2026, releasing a model designed to find and exploit security flaws to a small group of vetted customers and rating it “Critical” for cybersecurity under its Preparedness Framework. [2] The available excerpts verify parts of t
What should I do next in practice?
## Rating and intended use
Sources