Tencent AI Chief Warns Biology Is AI's Next Danger Zone as Hy4 Nears Release

Tencent's chief AI scientist, Vinces Yao Shunyu, used a Hong Kong science symposium to flag a hard problem: AI is getting good at biology, and that cuts both ways. The event was co-hosted by the Hong Kong Academy of Sciences and the Future Science Prize. Yao said frontier systems are advancing fast in biological understanding and could soon help researchers design drugs, identify treatments and crack complex scientific problems far quicker.

The catch, as Yao framed it, is that the same muscle that speeds up medicine also lowers the bar for harm. He compared biology to cybersecurity, where offensive and defensive tools are easier to tell apart — in biology, that line blurs. A model that helps scientists untangle proteins or build therapies, he warned, can just as easily assist someone trying to develop biological weapons or trigger a new pandemic.

Yao aimed much of his warning at open-weight models, the kind users can download, modify and run outside a company's walls. He said Chinese companies and open-source developers have to treat safety and security as their responsibility. His comments land as Tencent readies the weights and a safety system card for Hy4, its flagship model.

His prescription: test a model's biological capabilities before it ships, then build guardrails that block harmful assistance without choking off legitimate research. That puts biosecurity next to cybersecurity as a make-or-break test for frontier AI, and Yao says companies, regulators and research institutions will need to work closely together to keep the upside without opening the door to catastrophe.