Tencent AI chief: AI bioweapons are coming, but closing models isn't the answer
The Decoder · rss · 2026-11-09
Tencent chief AI scientist Vinces Yao Shunyu warns that AI will soon be able to develop bioweapons, but argues for a middle ground between open and closed models rather than blanket closure:
- Open models: safety filters can be easily stripped via abliteration
- Closed models: systems like Anthropic's Claude often block too aggressively
- Tencent's approach: for its upcoming open-weight model Hy4, dangerous hacking knowledge is filtered directly from the training data — safety at the source rather than at the filter layer
The position offers a third path in the open-vs-closed AI safety debate.
More from Models
- NeuroscienceBench debuts: 91 evals test if AI agents can analyze real neuroscience data — kenbwork · 2026-11-10
- Meta launches Muse Ember, its smallest model, claiming 3x cheaper than Haiku-class rivals — stablequan · 2026-11-10
- Gemini 4 Carbon beats Claude Opus 5.5: 23% lower cost per task, #1 on Intelligence Index — DeepLearningAI · 2026-11-10
- Gensyn live-trains open-8b: every step auditable on your laptop, overhead down to 1.3x — benfielding · 2026-11-10
- Meta's Muse Ember matches Claude Haiku 5.5 on Intelligence Index at a third of the cost — ArtificialAnlys · 2026-11-10
- Scale AI launches Muse Ember, a small model claiming Haiku-level intelligence at 1/3 the cost — alexandr_wang · 2026-11-10