OpenAI limits GPT-6 Cyber to Daybreak Red customers
First merged summary.
OpenAI has limited GPT-6 Cyber to Daybreak Red customers after tests it says found major Firefox and Linux flaws.
OpenAI has released GPT-6 Cyber, a cybersecurity model initially limited to approved Daybreak Red customers, with wider access planned. The company says it scored 48.5 percent on ExploitGym versus 32.1 percent for GPT-5.6 Cyber and 42.4 percent for GPT-6 Astra, and classifies it Critical. It reports 81 new high-severity Firefox flaws and working exploits for 948 Linux kernel vulnerabilities, including 588 previously unknown; The Decoder, Security Affairs and SecurityWeek also cite 1,724 reproduced kernel flaws. OpenAI says the model stayed in scope in 99.93 percent of honeypot tests. CyberScoop says the findings were patched, while The Register and CSO Online say Mozilla has started fixes and Linux reports are still under maintainer review.
- Access starts with vetted Daybreak Red API and Codex customers.
- ExploitGym score was 48.5 percent, above GPT-5.6 Cyber and GPT-6 Astra.
- OpenAI reports 81 previously unknown high-severity Firefox vulnerabilities.
- Working exploits covered 948 Linux kernel flaws, 588 previously unknown.
- Some outlets report 1,724 Linux kernel vulnerabilities were reproduced.
- OpenAI rates the model Critical and cites 99.93 percent scope compliance.
Timeline
dated events in the reporting · oldest first- Nov 4, 2026OpenAI released GPT-6 Cyber · Cyber Security News
What each outlet adds
beyond the shared facts- The VergeSays out-of-bounds behavior fell from about 5 percent to under 1 percent.
- CSO OnlineReports 10 novel techniques and under 25 percent remediation.
- The DecoderPuts the testing campaign cost at about $1.21 million.
- CyberScoopSays findings were patched and cites a Brotli remote code execution flaw.
- Help Net SecuritySays controlled tests let agents escape virtual environments.
- CVEs
- CVE-2026-111497
Coverage
oldest first · one row per article · UTC- · 7d ago
The Register · SecurityFirst report
OpenAI releases GPT-6 Cyber, rated Critical for cyber capability
Access starts with approved Daybreak Red customers after OpenAI said the model stayed in scope over 99 percent of the time.
- · 7d ago
The Verge · AI
OpenAI launches GPT-6 Cyber, rated Critical on its framework
OpenAI says out-of-bounds behavior fell below 1 percent from about 5 percent, and the ExploitGym score rose to 48.5 percent.
- · 7d ago
CyberScoop
OpenAI debuts GPT-6 Cyber for limited Daybreak Red customers
OpenAI says the model found 81 novel Firefox bugs and built exploits for 948 Linux kernel flaws, all since patched.
- · 7d ago
CSO Online
OpenAI's GPT-6 Cyber finds 81 unknown high-severity Firefox bugs and 588 kernel flaws
OpenAI claims 10 of the model's exploitation techniques were novel; its dashboard shows under 25% of found flaws remediated.
Vulnerabilities in this story
All| CVE | Vulnerability | CVSS | EPSS | Signals | Published |
|---|---|---|---|---|---|
| CVE-2026-111497 | Buffer overflow in Google Brotli decoder 1.0.3–1.2.0 | 9.2 | 83% | KEVPoC ×25 | KEV |