AI will soon master biology, unleashing new safety risks, Tencent AI chief warns

Tencent’s leading AI scientist, Vinces Yao Shunyu, warned that artificial intelligence will soon master biology, bringing both benefits and serious risks. As frontier AI systems are increasingly used in medicine and biology, they could accelerate drug discovery and treatments. However, experts also caution that malicious actors might use AI to create bioweapons or trigger the next pandemic, particularly if released as open-weight software.

Yao said at a Hong Kong science symposium that AI models and their biological abilities will soon reach a threshold where they can cause harm. He emphasized that all Chinese and open model companies must be responsible and take security concerns seriously. Tencent is exploring AI's potential to automate scientific research, with its latest flagship model, Hy4, ranking second-best in China.

Yao outlined plans to operate "autonomous science labs" using AI for complex experiments, such as inverse protein folding. While technical challenges can be addressed, the bigger hurdle lies in societal and organizational issues. Hy4's cyber hacking capabilities have been removed, but its biological capabilities remain below the high-risk threshold due to its inability to convert foundational biological knowledge into actionable protocols.

However, Yao cautioned about the complexity of managing AI's biological capabilities compared to its cyber counterparts. He stressed the importance of cautious release evaluations and collaborating with top security labs. Open-source versus closed-source AI models remains a contentious issue, with some experts arguing that open models can bolster defences, while others believe closed models are better for AI security.