AI Vendors Are Relearning the Oldest Rule in the Data Business

Software vendors are adding AI features that can use customer data to improve services for other accounts, raising questions about consent, fairness and what contributors receive in return. TechRadar’s Andy Frawley argues that vendors are becoming data companies without consistently setting clear rules for these exchanges.

Frawley compares the issue with data cooperatives formed by catalog retailers in the 1990s. Members pooled transaction histories to find potential buyers, with a straightforward bargain: contribute data to receive modeled prospects, and stop contributing to stop receiving them. Card fraud consortia use similar give-to-get terms.

HubSpot faced customer resistance after announcing plans to pool enrichment data across accounts, including business contact details, employer information and email deliverability signals. The data was intended to improve prospect records across the platform, and participation was initially on by default. Within days, HubSpot dropped the plan and said future enrichment features built on customer data would be opt-in. Frawley says sales and revenue operations teams objected because their work could improve tools available to competitors without a clear return.

Frawley cites McKinsey’s 2026 survey, which found that about 30% of organizations had reached a mature level in AI strategy, governance and controls. He says buyers are beginning to ask whether their data trains or enriches models for other customers, whether participation can be turned off, and what contributors receive. He also notes that removing a customer’s influence from a production model remains an unsolved problem for most vendors.

He argues that AI contracts should explain what data feeds shared models, what those models produce, what contributors receive and what happens to a model’s influence after a customer leaves.