AI Vendor Terms Risk Checklist
Tags: Frameworks
TL;DR
- Most AI tools ship with non-negotiable online terms—know the red lines: input rights, training data provenance, output ownership, and liability.
- Vendors often cap liability, exclude consequential damages, and limit indemnities—leaving output risks with customers.
Why it matters for HK marketers: Your campaigns and creative may rely on AI outputs—get the rights you need and avoid inheriting unbounded infringement risk.
Provisions to scrutinize
- Input rights: Limit use of your prompts/inputs to service delivery; no cross-customer training.
- Training data warranties: Lawful sourcing, license compliance, and no misuse of personal/sensitive data.
- Retention and deletion: Clarify what can be deleted and what cannot (e.g., learned parameters).
- Output rights: Secure ownership or exclusive rights to use, modify, distribute, and commercialize outputs.
- Consents and notices: Define who obtains and maintains permissions for third-party or personal data inputs.
Liability and indemnities
- Caps and carve-outs: Focus on exceptions (e.g., gross negligence, confidentiality, IP indemnity).
- Consequential damages: Negotiate exceptions for reputational harm, lost profits, and disruption.
- Indemnities: Push for IP indemnity; vendors often resist broad coverage for outputs based on your prompts.
Third-party data and scraping covenants
- ToS compliance: No reliance on data obtained in violation of site terms or licenses.
- Bot identity: Accurate user-agent strings linked to bot owner’s site.
- Anti-circumvention: No bypassing technical protections.
Many tools are offered only on non-negotiable website terms.
Vendors frequently cap liability and exclude consequential damages for AI outputs.
So what for marketers
Adopt a standard ToS checklist and escalation path; require amendments for high-risk use cases before integrating tools into production.
← Back to Knowledge Base