Tencent's chief AI scientist Yao Shunyu says AI's bio abilities will soon make it "possible to do harm" and Anthropic has a "strong argument" for closed models

AI models can now accelerate biological research and development, changing the pace and accessibility of scientific discovery.

Tencent's chief AI scientist, Vinces Yao Shunyu, warned that AI's growing capabilities in biology will soon allow for significant harm, comparable to its current abilities in cybersecurity. He stated that AI models are approaching a threshold where they can master biology, presenting dual-use risks similar to cyber capabilities, potentially enabling the creation of bioweapons or triggering pandemics. This warning comes as Tencent explores AI for scientific research, releasing its latest model, Hy4, which has biological capabilities below its high-risk threshold. Yao emphasized the need for responsibility and seriousness regarding safety concerns within the AI industry, particularly for open-weight models. He noted the narrowing gap between Chinese and US AI models and suggested a hybrid approach of both closed and open-source models might be optimal for balancing security and innovation.

The advancement of AI into biological domains signifies a critical inflection point, moving beyond theoretical potential to practical, albeit risky, applications. This progression necessitates a re-evaluation of safety protocols and ethical frameworks, as the dual-use nature of these capabilities means advancements in drug discovery could parallel advancements in bioweapon development. The debate between open-weight and closed-model approaches highlights the industry's struggle to balance innovation and security, with implications for global biosecurity and the democratization of powerful technologies.

Yao's assertion that societal and organizational challenges are more significant than technical ones underscores the need for robust governance and international cooperation. The potential for AI to lower barriers for malicious actors in biological attacks, as noted by China's ambassador for disarmament, suggests that policy and regulation must evolve rapidly alongside technological progress. The convergence of US and Chinese AI industries on frontier risk concerns, despite their differing model philosophies, indicates a shared understanding of the stakes involved in ensuring AI's safe integration into sensitive sectors like biotechnology.