LLMs spread errors in public services. Madrid’s open data portal offers a solution

New research from The Open Data Institute shows a big problem. Large language models often give wrong advice. They sound very sure of themselves, though. This is dangerous for public services like health care and benefits.

The study looked at more than 22,000 questions. It shows the risk for people who use these tools for important life choices. If you are just learning about a service, you might get bad information. This happens when the data is not clear or checked.

When data is not open, mistakes can spread. These “hallucinations” can hurt public services. But there is a good side to open data. It helps cities work better and stay open.

Madrid’s new Open Data Portal is a great example. It uses open-source technology and standard tools. This lets computers access data easily. It helps developers and citizens get value from public records. The platform shares information across city offices. It also follows European rules.

People who understand these systems can check facts better. This helps them in their daily lives. To learn how to use these standard systems, visit our guide on accessing open data. You will learn about tools like CKAN, DCAT, and OpenDataSoft.

Supporting open standards and clear licenses keeps public records safe. This protects everyone.

• What licence open data carries, and what you may do with it
• How to use an open-data portal’s API: CKAN, DCAT and OpenDataSoft
• How to file a FOIA request — and what to do when it is denied
• How to read a statistic before you quote it