Why ChatGPT Doesn't Belong in Government Offices and What Safe AI for Public Administration Looks Like (10 min)
ChatGPT is a quick fix for government agencies, but it also poses a huge risk. Leaks of sensitive data, hallucinations, and no integration with the city's internal systems. At Citymind, we've gone from a student project to a provider of AI solutions for Czech and Slovak municipalities over the past four years. We've implemented projects such as a European initiative for Prague 6. In this presentation, I'll demonstrate what an AI assistant for public officials, built on AI agents, looks like and what integrations are possible.
Artificial intelligence promises to speed up the work of offices, but in practice it runs into specific risks. In particular, small municipalities with a few employees need simpler and safer solutions than large "all‑in‑one" platforms. The talk focused on how to choose the right tool, comply with the GDPR, and avoid model errors. Slovakia has around 3 000 municipalities, most with fewer than one thousand inhabitants, and robust AI packages are often unnecessarily "heavy" for them. In practice, many offices then opt for a free version of a chatbot, which is a problem if citizens' data are sent outside a controlled environment. According to the speaker, it is also a risk when free services are used without clear contractual guarantees on data protection. Decisions should not be based only on popularity, but also on the tool's origin and its business model. With large models, answer quality can change over time and depends on the provider's priorities. In Europe, there is also growing pressure to prefer European and open‑source alternatives to avoid future restrictions in public administration.Why universal solutions often fail in small municipalities
