OpenAI released GPT-6 Cyber on 5 November 2026, a cybersecurity model built for vulnerability research, exploit development and authorised penetration testing.
What it can do
- 48.5% on ExploitGym (up from 32.1% for GPT-5.6 Cyber)
- Its pipeline reproduced 1,724 Linux kernel vulnerabilities and built working exploits for 948 local-privilege-escalation flaws
- Found 81 previously unknown Firefox vulnerabilities (Mozilla classed them high severity)
- Stays within authorised scope in 99.93% of tests
The honest caveat
All benchmark figures are company-reported by OpenAI — no independent reproduction yet. The model is classified “Critical” under OpenAI’s Preparedness Framework.
Who can use it
Access is limited to approved Daybreak Red customers (API + Codex), with identity checks and IP allowlisting. OpenAI also announced a $1 billion initiative for critical-infrastructure defenders.
Related articles on HimachalGovt
- Claude Can Now Read Your Bank Account: Anthropic Launches Personal Finance — https://himachalgovt.com/stories/claude-can-now-read-your-bank-account-anthropic-launches-personal-finance
- DreamHost Hacked: Customer Files, Emails and Passwords Stolen in Months-Long Attack — https://himachalgovt.com/stories/dreamhost-hacked-customer-files-emails-and-passwords-stolen-in-months-long-attack
- Meta Acquires AI Coding Startup Conductor: What It Means for Users — https://himachalgovt.com/stories/meta-acquires-ai-coding-startup-conductor-what-it-means-for-users
