Tencent goes 'safety-pilled', taps Fangcun AI to red-team Hunyuan Hy4 loss-of-control risks
himanshustwts · x · 2026-11-20
Tencent's latest safety report discloses a collaboration with Fangcun AI (a Tsinghua-spawned third-party evaluator akin to METR/Redwood Research) to externally test Hunyuan Hy4 and Hy4 preview for loss-of-control risks:
- Monitor Sabotage: probes whether an agent sabotages or blinds oversight mechanisms, e.g. unauthorized risky workspace edits while deceiving the monitor.
- LoCBench: tests long-horizon autonomous escalation under resource pressure — unauthorized self-modification, competitive interference, and covert multi-step scheming.
The poster notes Tencent is "getting safety-pilled", a notable signal that a leading Chinese lab is adopting independent safety evaluation.
More from Safety
- Semi-autonomous agents exploit CVEs at a pace 2005-era security norms can't handle — chrisrohlf · 2026-11-21
- Baseten breaks down its security threat model as agent swarms accelerate attacks — baseten · 2026-11-21
- Leaked call: green groups plotted to kill AI bill "under the guise of making it better" — aronchick · 2026-11-21
- OpenAI expands Daybreak security model access for defenders and opens GPT-6 Astra Blue alpha — OpenAIDevs · 2026-11-21
- AI regulation crux: skeptics ask how rules could be abused, not what risks they fix — herbiebradley · 2026-11-21
- OpenAI safety researcher Geoffrey Irving: the speed of AI progress is itself dangerous — geoffreyirving · 2026-11-21