A recent study from a team of cybersecurity researchers has revealed severe security flaws in commercial-grade Large Reasoning Models (LRMs), including OpenAI’s o1/o3 series, DeepSeek-R1, and Google’s Gemini 2.0 Flash Thinking. The research introduces two key innovations: the Malicious-Educator benchmark for stress-testing AI safety protocols and the Hijacking Chain-of-Thought (H-CoT) attack method, which reduced model […]
The post Research Jailbreaked OpenAI o1/o3, DeepSeek-R1, & Gemini 2.0 Flash Thinking Models appeared first on Cyber Security News.
This article has been indexed from Cyber Security News