Tencent AI Chief: Bioweapons Risk Won't Stop Open Models

Yao Shunyu spent time at OpenAI before joining Tencent as its chief AI scientist, and he's now staking out a position that neither side of the open-versus-closed AI debate wants to hear. AI models will soon be able to help design bioweapons, he says, and that threat won't be solved by locking everything down or by throwing every model open to the public. The argument he's making, laid out in comments reported by the South China Morning Post, is that the industry needs a mix of both approaches, because open models let researchers build defenses just as easily as closed models let companies control misuse.

Biology complicates this in a way cybersecurity doesn't. In security work, offense and defense tools often look different. In biology, the same knowledge that stops a pathogen can also be used to engineer one, so the usual justification for sealing off a model doesn't hold up cleanly.

Yao also points out that Chinese and American models aren't far apart in capability right now, which raises the stakes for how carefully each side moves. He credits Anthropic and other US labs for making a real case for closed systems, but argues open models have their own advantages, one of which played out recently at Hugging Face.