A new Chinese artificial intelligence model is raising cybersecurity fears due to its advanced capabilities, which appear to be on par with Claude Opus 4.8 and OpenAI's GPT-5.5, Axios reported. The model is also about half the cost of its U.S.-based counterparts.
"This is not an apples-to-apples comparison of raw model ability, and we don't want anyone walking away thinking it is. Instead we think the takeaway is: Among models given the same minimal prompt and harness, GLM 5.2 a open-weight model, ⅙ the cost of a frontier LLM beat Claude Code at a genuinely difficult security research task," an analysis by Semgrep said.
A Graphistry analysis of GLM-5.2 stated that it might be an "illegal distillation of both GPT-5.5 and Opus 4.8." The analysis further noted that "Anthropic reported several months ago that Chinese-origin model companies are performing distillation attacks to steal their model weights, so high correlation scores and similar incorrect answers are noteworthy."
The open-weight nature of GLM-5.2 also means that it could create cybersecurity issues. Axios reported that GLM-5.2 can be downloaded and modified. That means that safety controls can be removed and that it can be altered for specific tasks independently.
The fear is that hackers could take GLM-5.2, which performs well in comparison with some of the most advanced U.S.-based AI models, and use it for nefarious purposes.
Jason Baker, managing security consultant at GuidePoint Security, told Axios that Russian-language forums already included discussions about the ease of taking GLM-5.2 and transforming it for use in hacking.
Travis Lanham, CTO and founder of Armadin, added that GLM-5.2 might allow hackers to personalize their attacks and find creative ways to beat systems.
"An attacker can run it locally without safety guardrails, fine-tune it against their specific targets, and operate with zero visibility to any provider or defender," Lanham told the outlet.
