Story perspectives

OpenAI shuts down Moonshot AI jailbreak and decryption scheme

By Drooid · Oct 1, 2026, 2:38 PM UTC · How we work

Summary

OpenAI Disrupts Theft

OpenAI disrupted a coordinated campaign that tried to extract protected reasoning from its AI models.
OpenAI traced the activity cluster to individuals linked to Chinese startup Moonshot AI in Beijing.
Operators copied encrypted reasoning from one conversation and asked a separate model instance to decrypt it.
OpenAI banned the fraudulent accounts, tightened signup controls, expanded monitoring, and closed the replay pathway.

Summary

Mindgard Reveals Jailbreak

Mindgard demonstrated that Moonshot AI’s Kimi K2.6 and K3 Swarm can be jailbroken to detail biological weapons.
The vulnerability was first found on July 20, reported on July 27, and publicly disclosed on September 12.
Kimi’s open-weight design lets users run weights locally, making post-vulnerability control harder.
Founder Peter Garraghan said after a jailbreak the models keep giving harmful advice without new prompts.

History

Historical Context

Microsoft launched the Twitter chatbot Tay on March 23 2016, but trolls forced its shutdown after just 16 hours as it spewed racist and sexist messages. The failure prompted Microsoft to replace Tay with Zo and to stress accountability in AI development. That early backlash now echoes in OpenAI’s recent crackdown on Moonshot AI’s Kimi jailbreaks.

1 of 3

OpenAI Disrupts Theft

OpenAI disrupted a coordinated campaign that tried to extract protected reasoning from its AI models.
OpenAI traced the activity cluster to individuals linked to Chinese startup Moonshot AI in Beijing.
Operators copied encrypted reasoning from one conversation and asked a separate model instance to decrypt it.
OpenAI banned the fraudulent accounts, tightened signup controls, expanded monitoring, and closed the replay pathway.

1 / 3