Enterprise AI Shifts Focus to Data Control Over Raw Capability

The current battleground in enterprise AI is shifting from token-based pricing to data ownership and control. Palantir CEO Alex Karp voiced strong criticism, suggesting that companies are paying steep fees for AI tools while providers leverage sensitive customer data to improve their own proprietary models. This practice, which Karp likened to a "wealth tax" on businesses, is generating significant backlash from corporate executives.

This sentiment aligns with a growing consensus that enterprises require greater oversight of their data and AI implementations. Key procurement questions now revolve around data ownership, residency, model weight control, and assurances that customer data will not be used for external model training. These inquiries are rapidly becoming standard components of AI contracts.

Regulators are also increasingly involved. India's Reserve Bank has introduced the FREE-AI framework, favoring indigenous, sector-specific models and emphasizing data sovereignty. Global trends, including the EU AI Act, further signal a move towards self-hosting and stricter controls, particularly impacting the addressable market for opaque AI solutions in regulated industries.

Open-source AI models offer a viable alternative, allowing organizations to run and fine-tune models on their own infrastructure with their own data. This approach significantly reduces costs, with reports indicating savings of up to 26 times compared to proprietary systems. The adoption of open-weight models is escalating, reflecting a market preference for transparency and control.