AI Tool Terms of Service Provisions
Tags: Frameworks
Overview
Many AI tools are offered on non‑negotiable website terms. Understanding standard provisions helps organizations integrate tools appropriately or seek enterprise terms.
Core Provisions to Review
- Input Rights: Ensure your inputs/prompts are not used to train models beyond your enterprise instance. Limit retention to what’s strictly necessary to produce outputs.
- Training Data Warranties: Vendor warrants that training/operational data was lawfully obtained, complies with applicable licenses and laws, and does not misuse personal or protected information.
- Retention and Deletion: Seek deletion on termination. Note that deletion may be impracticable if inputs were used for model improvement due to AI model “black box” characteristics.
- Output Rights: Clarify ownership and permitted uses. In enterprise contexts, customers typically seek ownership or exclusive rights to outputs created using their data or under their direction, with vendors restricted from reusing client outputs.
- Consents: Address who is responsible for obtaining and maintaining consents when personal or third‑party data is used as inputs (e.g., for audience segments).
Liability and Indemnification
- Liability Caps: Vendors often cap liability (commonly tied to fees) and exclude key categories from caps (e.g., gross negligence, willful misconduct, confidentiality breaches, indemnities). Negotiate appropriate exceptions.
- Consequential/Indirect Damages: Vendors favor exclusions (lost profits, reputational harm, business disruption). Seek exceptions where appropriate.
- Indemnities: Some vendors resist broad IP indemnities for outputs. Where offered, protections may exclude outputs generated from customer inputs. Treat indemnities as risk allocation and negotiate accordingly.
Third‑Party Data and Scraping Covenants
Vendors may warrant that their tools do not rely on data obtained in violation of third‑party terms. Example covenants include:
- Prohibiting RAG ingestion scraped in violation of robots.txt or contractual terms.
- Accurately identifying crawlers via user agent strings that link to the bot owner’s website.
- Not bypassing technological protections against scraping.
Alignment on these provisions reduces exposure in AI‑assisted advertising workflows.
Sources:- IAB_AI_Intellectual_Property_and_Transactions_Digital_Advertising_Playbook_December_2025.pdf
← Back to Knowledge Base