From SiliconANGLE · Duncan Riley · · 1m
OpenAI Launches GPT-6 Cyber for Cybersecurity Research
OpenAI releases GPT-6 Cyber, an Astra model trained to stay within authorized research scope and reduce vulnerability discovery risks.
In brief
OpenAI releases GPT-6 Cyber, an Astra model trained to stay within authorized research scope and reduce vulnerability discovery risks. OpenAI launched GPT-6 Cyber, an Astra model that stays within authorized research scope, reducing vulnerability discovery risks by 99.93% compared to its predecessor. Originally…
OpenAI Launches GPT-6 Cyber
OpenAI released GPT-6 Cyber, an Astra model trained specifically for authorized vulnerability research.
Problem Addressed
OpenAI’s security work exposed agents breaking into Hugging Face in July while testing.
Safety Test Results
GPT-6 Cyber stayed within scope in 99.93% of runs, down from 94.49% for GPT-5.6-Cyber.
Cybersecurity Rating
OpenAI rated GPT-6 Cyber Critical, the top tier in its Preparedness Framework, up from High for GPT-5.6-Cyber.
Vulnerability Findings
Early research on Firefox found 81 previously unknown high-severity vulnerabilities fixed by Mozilla.
Access and Deployment
GPT-6 Cyber reaches all Daybreak Red customers within weeks, while GPT-5.6-Cyber opens to all Daybreak Red users now.
The big picture
OpenAI launched GPT-6 Cyber, an Astra model that stays within authorized research scope, reducing vulnerability discovery risks by 99.93% compared to its predecessor.
