CData launches Connect AI Gateway for agent governance

CData's move extends its earlier Connect AI platform, which focused on giving AI tools governed access to enterprise systems, into a product meant to oversee the whole request: the initial prompt, model selection, data retrieval and any resulting action. That broader scope is what the company says would let administrators investigate an agent's behavior when an answer is wrong or an action is blocked.

The design leans on CData's existing connectors to hundreds of business applications and databases. Those connections are what let the gateway enforce a person's permissions when an agent retrieves data or acts on that person's behalf, with policies governing access to specific rows and columns. A context engine adds business definitions and less formal knowledge, held in a context graph outside individual AI models so an organization can reuse it when it changes models.

CData is careful about where the product stops. Executives said the gateway can import business definitions from analytics products but is not meant to replace a data catalog or semantic layer, and imported definitions from data catalogs are planned, not present. The open question is governance of shared knowledge: a correction useful to one employee could mislead others, so CData says human review decides what becomes shared context. How smoothly that approval process runs is likely to shape how much value customers get from the platform.