Tencent's AI chief says AI bioweapons are coming, but closing off models isn't the answer
Reported by The Decoder1 source ↓
Tencent chief AI scientist Vinces Yao Shunyu warns that AI will soon be capable of developing bioweapons. He argues that neither fully open nor fully closed models are sufficient on their own.
Key points
- Yao says open models can help defenders, but biological offensive and defensive capabilities are hard to separate
- Yao says the gap between Chinese and US models is small and urges the industry to proceed carefully
- Hugging Face used open model GLM-5.2 for forensic analysis after closed models refused the task
- Tencent plans to release open-weight model Hy4 with a safety report in the coming days
Details
The Decoder reports that Yao, Tencent’s chief AI scientist and a former OpenAI researcher, is arguing for a mix of open and closed models. He says open models can also support defensive work, but warns that offensive and defensive capabilities are harder to separate in biology than in cybersecurity. Yao says the gap between Chinese and US models is not large.
Yao cites a Hugging Face security incident in which the platform used Chinese model GLM-5.2 to analyze attack logs after closed US models refused the task. The Decoder notes that GLM-5.2 was used only for forensic analysis; Hugging Face stopped the attack by patching the vulnerability, rebuilding affected servers and changing credentials. Hugging Face said the incident was not an argument against safety filters on hosted models.
Tencent plans to release its Hy4 model as an open-weight model, accompanied by a safety report, in the coming days. The company says Hy4’s biological capabilities are below the high-risk threshold and that it removed the model’s hacking capabilities by filtering training data. The article says researchers found this approach persisted after a technique that removes a model’s refusal behaviors.
Written from coverage as of
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh