Does Kimi train on your data?
Kimi says it uses what you type and what it answers, de-identified, to optimise its models by default. To stop it you write to customer service and it stops.
"On the premise that the information has been processed with secure encryption technology, strictly de-identified, and cannot be used to re-identify any specific individual, we will use the content you input and output to optimize our models, or use your access to and use of our products for data analysis, so as to enhance your service experience."
- 6 Sept 2026
- All accounts (kimi.com)
- global
- Saved by hand — not re-read automatically
- BC737080…8051
How we read it: The policy names both the use and the route in consecutive paragraphs. The use: "we will use the content you input and output to optimize our models", on the stated premise that it is encrypted and "strictly de-identified". The route: "If you do not want your content to be used for model training, you may contact our customer service through the contact details set out in Section 11 of this Policy and make a request. After verifying your identity, we will stop using your content for model training." A request rather than a setting, so a D, at the lower end of the band for the de-identification and the plainly stated route. Saved by hand because the page loads behind the app's own shell.
What each plan can refuse
The same questions, asked of every plan we could identify. Every cell comes from a quoted clause or a documented search - see the evidence above.
- Contact support
- Not stated
- 52 / 100
- Yes, stated
- Unknown
How to opt out of Kimi AI training
Plan by plan, from what Kimi's own policy says about refusing. Where it publishes the steps, they're listed here with the limits they carry. Where there's nothing to switch off, or nothing graded, this says so instead of inventing a menu path.
All accounts (kimi.com)
Yes, but not self-serve: Kimi makes you contact support.
What this policy has done since we started watching
The record starts here. Kimi (Moonshot AI)'s policy was first captured on 6 Sept 2026, and there's no earlier copy to compare it against, so this verdict has no history behind it yet. From now on every re-read is diffed against that copy, and anything that moves appears here.
Everything we read
Vendor response
We haven't yet written to Kimi (Moonshot AI) about this verdict, and they haven't disputed, clarified or announced a change to any clause quoted above. When either happens it's published here, word for word.
Report a wrong clause