AI raises two IP questions for buyers: who owns what the AI produces, and who pays if the AI or its output infringes someone else's rights.
Ownership of outputs
Most business AI terms say that, as between you and the vendor, you own (or the vendor assigns to you) the outputs. That settles the contract position. It doesn't settle whether the output is protectable: in several jurisdictions, including the U.S., copyright requires human authorship, so purely AI-generated material may have limited protection. For important assets (brand, code, product content), keep a record of human contribution.
Infringement risk and indemnities
Models are trained on large datasets, and there is ongoing litigation about training and outputs. Major providers have responded with IP indemnities for paying business customers:
- Microsoft announced its Copilot Copyright Commitment in September 2023, later the Customer Copyright Commitment: it will defend paying commercial customers against copyright claims over its Copilot services and their output, and pay adverse judgments or settlements, if the customer used the built-in guardrails and content filters.
- Google Cloud announced in October 2023 a two-part generative AI indemnity covering its use of training data and the generated output of certain services.
Read the conditions
Indemnities come with conditions: paid tiers only, use of safety filters, no intentional infringement, no modified models, limits on which products are covered and sometimes caps. Check that your actual use fits, and that the indemnity survives termination.
Your own data and inputs
You warrant that you have the right to use what you upload. Uploading third-party content you don't have rights to can void an indemnity.