What it means for a UK law firm
Model training on client data is a confidentiality risk because information used in training can become part of the model, and in some circumstances models can reproduce training data in answers to other users.
Model training on client data usually depends on the product tier and contract, not the brand. Consumer and business versions of the same AI product often have different default terms, so a firm should check the current terms for the specific product and account type it uses.
Example
Illustrative scenario
Two colleagues use the same AI product, one through a personal free account and one through the firm's business account. The firm's contract excludes training on its data; the personal account's default settings may not. Only the business account should be used for client work.
How Chancery handles it
Chancery only works with AI providers who guarantee that a firm's data is not used to train or improve their models; Part 1 of our series on using AI safely explains why this matters.