FreeLLMOpenAI unveils GPT‑6 Cyber, a Critical‑level AI for vulnerability research
The new model scores 48.5% on ExploitGym, finds dozens of bugs, and ships to Daybreak Red users with tighter safety limits.
OpenAI announced GPT‑6 Cyber, the latest version of its cybersecurity model, and placed it in the Critical tier of its Preparedness Framework.
In internal tests the model stayed within bounds for more than 99% of queries, a drop from roughly 5% out‑of‑bounds behavior seen with GPT‑5.6 Cyber.
The model helped uncover 81 previously unknown vulnerabilities in Firefox and was used to scan the Linux kernel, identifying 948 potential issues.
OpenAI is rolling GPT‑6 Cyber out to participants in the Daybreak Red program and plans to extend access to all program customers in the coming weeks.
O que isso significa para você
The tighter safety controls mean defenders can use a powerful AI tool without as much risk of it being repurposed for attacks. For users who rely on free or locally hosted models, the launch highlights how frontier models are increasingly gated behind specialized programs, raising the performance gap between open tools and proprietary AI security solutions.
Fontes
- OpenAI released a new cybersecurity model. Is that enough?, TechCrunch, 5 Nov 2026
- OpenAI says its new cybersecurity model is better at not hacking other companies, The Verge, 5 Nov 2026
Escrito automaticamente a partir das fontes acima, depois verificado pelo nosso código: todo número e nome deve aparecer nelas, e nenhuma frase pode copiar uma fonte. Encontrou um erro? Reporte‑o.
Qual IA você pode usar grátis hoje?
Todo modelo de IA gratuito, classificado por votos humanos cegos, com a memória e velocidade que ele precisa na sua própria GPU.