Working with AI Providers Without Betting the Firm
Details
Model, provider, and hosting location are three different contracting facts, and most firms have signed up for all three by letting a staff member create an account. This session is the owner-level view of working with AI providers: what leaves your boundary, what never should, and how to keep the right to change your mind.
In 45 minutes:
- The three nouns owners mix up: the model, the provider API, and where the bytes are processed
- PII minimization in practice: stripping and masking before a model sees the file, and what logging still exists afterward
- The vendor-intake checklist: data terms, subprocessors, retention, training use, and the right to swap models
- Claims you can make to clients versus claims you cannot
Then open discussion: bring a vendor agreement you're unsure about (names redacted welcome).
