San FranciscoOpenAI says GPT-6 Cyber found 948 Linux flaws.
A model that finds bugs at scale serves attackers too, so OpenAI limits access.
It compiles the software and runs regression tests to confirm each patch works before suggesting it.
Only verified security teams, infrastructure operators and certified researchers can use it, and OpenAI decides who qualifies.
How each outlet framed it
drawn from 70+ reports worldwide
- Mashable ME leans favorable
- echoes OpenAI launch copy: 948 Linux flaws found in pre-deployment testing, plus safeguards against offensive reuse
Sources: Mashable ME