Data governancepartial
What happens to your data: training-on-customer-data defaults, retention windows, residency options, and the enterprise-versus-consumer terms gap. A legal property, not a capability — it survives every model generation.
Xiaomi's own text now confirms the no-training default, and states it categorically rather than as a configurable option: submitted content is not used for training or any other purpose. Residency is disclosed (Netherlands and Singapore) with EU Standard Contractual Clauses and a named EEA controller. What holds this short of strong is retention: the policy commits only to keeping data for 'the period necessary', with no defined window and no zero-retention option — so the 30-day prompt-logging figure carried by secondary reporting remains unconfirmed by the provider. Self-hosting the MIT weights is deployer-controlled.
Receipts (3)
- Xiaomi MiMo Privacy Policy §3.1 — training use
“Xiaomi will not use the content you provide for model training or any other purposes.”
- Xiaomi MiMo Privacy Policy §10 — data residency
“Currently, Xiaomi has data centers in the Netherlands, and Singapore.”
- Xiaomi MiMo Privacy Policy §6 — retention (undefined window)
“As a general rule, we retain personal information for the period necessary for the purposes described in this Privacy Policy, or as required by applicable law.”
The gap that keeps this partial rather than strong: no numeric retention window and no zero-retention control.