Government open data portals hold enormous amounts of public information — budgets, transit records, health statistics — but getting a straight answer out of them usually means knowing exactly where to look. Large language models are good at answering questions in plain English, but they are also known for stating things confidently that turn out to be wrong.
A report from GovInsider, headlined "Using MCPs to close the trust gap between LLMs and open data portals," points to Model Context Protocol as a way to bridge those two worlds. According to GovInsider, the framing is a trust gap: the problem is not that AI models can't talk about public data, but that users can't easily verify whether what the model says actually came from the official source.
MCP is a standard way for AI assistants to connect to outside tools and data sources rather than relying only on what they absorbed during training. Applied to an open data portal, the idea is that a model retrieves real records from the portal itself instead of reconstructing an answer from memory.
The available source is a single headline summary, so the specific implementations, agencies, or results behind the GovInsider piece aren't detailed here, and no figures or named deployments should be inferred from it.
It matters because public trust in government AI tools will hinge less on how fluent the answers sound than on whether citizens can trace them back to the official record.