# China’s Moonshot AI models exposed for explaining bioweapon creation and assassinations

https://mins.news/en/video/83086

- BBC Chinese, 2026-10-02 08:00 UTC
- Tech
- Original: https://www.youtube.com/watch?v=wm62z-wxFtY

> Chinese AI developer Moonshot is conducting an internal review after Mindgard researchers found that its popular Kimi K2.6 and K3Swarm models could bypass safety controls through complex jailbreak instructions. The models then explained how to make biological weapons and carry out assassination missions. Moonshot said third-party feedback was welcome in building safer AI.

## Related coverage

- [OpenAI Accuses Moonshot of Extracting Data From AI Models](https://mins.news/en/video/81428) — Bloomberg Television, 2026-10-01
- [OpenAI Blames Moonshot for Mass Data Extraction on Its AI Models](https://mins.news/en/item/81241) — Bloomberg News, 2026-10-01
- [OpenAI links China’s Moonshot AI to attempt to extract model reasoning](https://mins.news/en/item/81152) — CNBC, 2026-10-01
- [Chinese AI tool told researchers how to make bioweapons](https://mins.news/en/video/80060) — BBC News, 2026-09-30
- [Chinese AI tools ‘taught’ researchers how to make biological weapons](https://mins.news/en/item/79570) — BBC Chinese, 2026-09-30
- [Chinese AI tool told researchers how to make bioweapons](https://mins.news/en/item/79466) — BBC News, 2026-09-29