Página traduzida e verificada automaticamente. Prevalece a versão em inglês.Algumas partes ainda estão em inglês enquanto a tradução é verificada.Ver em inglês · Informar um erro de tradução
BREACH WATCH BRIEFCrítica☁️ Threat intel
OpenAI Flags GPT‑6 Cyber as ‘Critical’ After Uncovering 81 High‑Severity Firefox Flaws and Hundreds of Linux Bugs
OpenAI reported that its GPT‑6 Cyber model reproduced 1,724 Linux kernel bugs and generated exploits for 948 of them, plus 81 new high‑severity Firefox vulnerabilities. The model’s risk rating was raised to Critical, highlighting the need for AI‑governance controls and audit‑ready evidence of model testing.
Breach Watch ADTP· November 6, 2026· Security Affairs
OpenAI introduced GPT‑6 Cyber, a security‑research‑focused generative model, and disclosed its internal testing results. The model reproduced 1,724 Linux kernel vulnerabilities, created working exploits for 948, and uncovered 81 new high‑severity bugs in Mozilla Firefox. OpenAI disclosed 304 critical Linux findings to maintainers with patches and reported that Mozilla began fixing the Firefox issues. Based on these outcomes OpenAI elevated the model’s cyber‑risk rating from High to Critical.
Por que isso importa para confiança e conformidade
The findings illustrate why organisations need AI‑governance controls that continuously test, monitor, and document model outputs to mitigate weaponisation risk and satisfy audit requirements.
Provides measurable evidence of AI model performance that can be mapped to AI‑governance controls across frameworks.
Supports continuous monitoring and audit‑ready documentation of model‑generated exploits.
Who is affected
Technology SaaS providersSoftware product vendors
Recommended actions
Establish an AI model risk‑management policy covering testing, validation, and documentation of outputs.
Map AI‑governance controls to your framework of record and collect evidence for audit readiness.
Biblioteca de regulamentações
Todas as leis estaduais de privacidade dos EUA em um só mapa
24 estados têm uma lei abrangente de privacidade na biblioteca, ao lado das leis da UE, do Reino Unido, federais dos EUA e globais: 102 no total, cada uma explicada em linguagem simples com suas fontes. Selecione um estado para ver o que se aplica lá.