Tencent AI chief warns models will soon be capable of biological harm

Vinces Yao Shunyu, Tencent Holdings' chief AI scientist, says artificial intelligence is closing in on a dangerous milestone. During a rare public appearance on Saturday, the former OpenAI researcher said models will soon be capable enough at biology to cause genuine harm. He was speaking at a science symposium co-hosted by the Hong Kong Academy of Sciences and the Future Science Prize.

His warning was blunt. The cyber abilities of AI models — and, before long, their biological abilities — will hit a threshold where real-world damage becomes possible, he said. Yao framed the moment as a question of when, not if.

Frontier AI systems are already being pointed at medicine and biology, where the upside is hard to ignore. Faster drug discovery and new treatments for stubborn diseases are the pitch. That same horsepower, though, is what worries the people building it.

Experts have flagged a darker scenario: bad actors using AI to cook up bioweapons or trigger the next pandemic. The risk spikes when models ship as open-weight software, according to the South China Morning Post, since the weights can be downloaded and used far from any lab's oversight.

Yao's résumé gives the warning extra weight. He arrived at Tencent from OpenAI, one of the most closely watched talent moves in the industry, and now leads AI research at one of China's biggest tech companies. His message lands as the field races to push models into sensitive scientific territory — benefits and hazards arriving together.