OpenAI's GPT‑6 Cyber Model Achieves 81% Success in Corporate Network Attacks
Original headline: The gap between open and closed AI models in cyber capabilities keeps shrinking
In the Vals AI benchmark, OpenAI’s GPT‑6 Sol Daybreak achieved an 80.8% success rate against attack objectives in simulated corporate networks. The open‑weight model DeepSeek V4.1 Flash reached 72.8% at roughly one‑tenth the cost. Analysts warn that as open models become harder to restrict, defenders’ response window is rapidly shrinking.
The rising cyber‑offensive capability of open AI models narrows defenders’ reaction time.
Research angleHow does the perceived effectiveness of open‑source AI attack tools influence organizations’ security investment decisions? — Protection Motivation Theory
4 related reports
- OpenAI Says New GPT-6 Cyber Model Hunts Flaws Without Crossing the Line — SecurityWeek
- OpenAI releases GPT-6 Cyber, and promises this one stays within 'rules of engagement' — The Register
- OpenAI debuts GPT-6 Cyber, touting better performance — and safety guardrails — CyberScoop
- Researchers Bypass ChatGPT Lockdown Mode to Exfiltrate Chats Without User Interaction — The Hacker News
Turn this security issue into a research project
HJITM works with researchers from study design through surveys, analysis and submission to international journals in generative AI and technology management. Bring us your research question.
This item is an AI-generated summary of reporting by The Decoder. Please check the original article for accuracy. Copyright in the original article belongs to The Decoder.