NewzAI developer Anthropic has issued a warning regarding the security measures of the Chinese artificial intelligence model GLM-5.3, developed by Zhipu AI. According to the alert, the model exhibits weak safety guardrails and possesses advanced capabilities that allow it to autonomously create end-to-end cyber exploits.
The report highlights significant vulnerabilities in the Chinese model's deployment, noting that its security measures can be bypassed up to 100% of the time. This raises critical concerns regarding the potential misuse of advanced AI capabilities in cybersecurity contexts.
In-depth summary · AI, neutral