liveBook · Manning

No-training option

As explained in AI Governance

A no-training option is a setting or contractual commitment that prevents customer-provided data from being used to train or improve shared models.

A no-training option is a setting or contractual commitment that prevents customer inputs from being used to train shared models. It can limit reuse risk when confidential information enters a provider environment, but does not by itself settle ownership of resulting model weights or eliminate breach exposure. section 6.6.2

Check provider terms for risks to:

  • Attorney-client privilege, if inputs may be collected, trained on, or disclosed. section 6.6.2
  • Trade-secret status, if confidentiality protections are absent or unenforceable. section 6.6.2
  • Patentability, if sharing an unfiled invention amounts to public disclosure. section 6.6.2

Fine-tuning with proprietary manuals

ClaimAssist was fine-tuned using proprietary repair guides, while its contracts left ownership of the resulting weights unclear. A no-training commitment addresses a different risk: reuse of customer inputs to train shared models. section 6.6.2

How AI Governance covers it

How deep it goes
Traces training and disclosure terms to privilege, trade-secret, patentability, and model-ownership risks, without detailing implementation.
Examples
Uses ClaimAssist’s proprietary repair manuals and discusses court disputes involving AI-platform disclosures.
What you can do after
Evaluate provider terms for training and disclosure clauses that could threaten confidentiality or intellectual-property protections.
How it connects
Connects provider training terms with attorney-client privilege, trade secrets, patentability, fine-tuning, and model-weight ownership.

This explainer was generated by Manning's knowledge graph from the text of AI Governance. The section links point to the parts of the book it draws on.