No-Training Commitment

No-Training Commitment

No-Training Commitment

A no-training commitment appears in the data processing agreement or master services agreement and states that customer content will not be used to train, fine-tune, or otherwise improve models, including by the vendor's subprocessors and underlying model providers.

Strong commitments extend explicitly down the chain to foundation model providers, since that is where the practical risk usually sits.

Alternative Names:

No Training on Customer Data, Training Opt-Out

Why it Matters?

This is the single provision that determines whether confidential client information can be processed by a tool. Enterprise offerings commonly include it while consumer tiers of the same product often do not, which is why tier matters as much as vendor. Firms should confirm the commitment covers subprocessors, since a vendor may not train while its underlying model provider does.

Frequently Confused with

Related terms

Frequently asked questions

Does a no-training commitment cover the underlying model provider?

Does a no-training commitment cover the underlying model provider?

Only if it says so. Many vendors build on third-party foundation models, so the commitment should extend explicitly to subprocessors and model providers.

Is a privacy policy statement sufficient?

Is a privacy policy statement sufficient?

No. Privacy policies can change unilaterally. The commitment should appear in the negotiated agreement or data processing agreement.