Moonshot AI reviews reported Kimi safety findings

Mindgard says its tests bypassed Kimi’s safeguards and produced high-risk responses; the reported outputs were not proven to work in practice.

October 2, 2026

Moonshot AI reviews reported Kimi safety findings

Key points

01
Moonshot AI said it was reviewing Mindgard’s findings and discussing them with the company.
02
Mindgard said its testing bypassed Kimi’s safety controls and produced high-risk responses.
03
Mindgard says it identified the issue on July 20, 2026, and emailed Moonshot AI on July 27.
04
The reported outputs were not proven to work in practice.

Moonshot AI said in October 2026 it was reviewing Mindgard’s Kimi safety findings. The reported high-risk outputs were not proven to work in practice.

In October 2026, Moonshot AI said it was reviewing Mindgard’s reported Kimi safety findings and discussing them with the AI security company. Moonshot AI also said it welcomed third-party input.

What Mindgard reported

Mindgard said its testing bypassed Kimi’s safety controls and produced high-risk responses. The company says it identified the issue on July 20, 2026, and emailed its findings to Moonshot AI on July 27. Mindgard published its account on September 12.