Data Marketplaces & Intermediaries

Report: enterprise SaaS contracts are quietly becoming AI training licenses

Source: PYMNTS · Jun 1, 2026

A PYMNTS analysis of enterprise SaaS contracts finds that 92% of AI vendor agreements claim data-usage rights that extend beyond simply delivering the contracted service — meaning the vendor can typically use customer data for other purposes, including AI training, unless the contract specifically carves that out. Only 33% of those contracts offer the customer third-party IP protection, and just 17% commit to following applicable law on how that data gets used. The piece cites Figma's November 2025 lawsuit over alleged unauthorized use of customer designs for AI training, and notes Adobe's public commitment not to train on customer data after facing similar backlash.

This is the regulatory and contractual environment our entire pitch operates inside: most enterprise software today obtains training rights to customer data by default, buried in terms most customers never read closely, rather than through explicit, informed consent. That's the opposite of how we source interview data — every dataset we license is built on documented, specific consent from the actual people in it, not an omnibus contract clause. As legal and PR exposure around default-opt-in data licensing keeps surfacing (Figma being the latest, highest-profile example), the contrast between blanket contractual rights-grabs and rights-cleared, consented data sourcing becomes a sharper part of the pitch, not a softer one.

Key Points

• 92% of AI vendor contracts claim data-usage rights beyond service delivery
• Only 33% of those contracts offer third-party IP protection to the customer
• Only 17% commit to following applicable law on data use
• Figma sued in November 2025 over alleged use of customer designs to train AI; Adobe committed not to train on customer data after backlash