Tencent AI chief Shunyu Yao warns model bio abilities will soon reach a threshold to do harm

AI models will soon be powerful enough to master biology, bringing both benefits and serious risks, Tencent chief AI scientist Shunyu Yao said on Saturday, the South China Morning Post reports. "The cyber abilities of models and soon the bio ability of models will reach a threshold where it's actually possible to do harm," Yao, a former OpenAI researcher, said at a science symposium co-hosted by the Hong Kong Academy of Sciences and the Future Science Prize.

He said Chinese and other open-model companies need to be "very responsible and very serious" about these security and safety concerns, noting that the gap between Chinese models and their US counterparts is not large. He also outlined Tencent's plans for "autonomous science labs" that use AI to run complex experiments end to end, such as inverse protein folding.

Tencent says it removed cyber hacking capabilities from its Hy4 model, released on Wednesday, by filtering training data, and that the model's biological capabilities remain "well below" its high-risk threshold. Yao said US companies such as Anthropic have made a "strong argument" that frontier models should be closed.