Azerbaijan says it is developing Azerbaijani-language AI models for state services, arguing that systems built for much larger language markets can give shorter, lower-quality answers in smaller ones. The plan brings language access and cultural context into a practical question: who gets useful AI when public services move online?
Watch Desk analysis
What happened
Deputy Minister of Digital Development and Transport Rashad Hasanov said the models are being developed for state services during a discussion at the World Forum on Intercultural Dialogue in Baku, Trend.Az reports. He argued that responses in languages spoken by relatively few people can be incomplete and may drift towards English, even when they are not simply incorrect.
Azerbaijan has already introduced speech-to-text conversion and its first agent-based services in the my.gov application, which connects citizens with the state. The government plans to expand the number of services there by the end of the year. Hasanov also said a programme is planned to train more than 25,000 civil servants from 275 institutions in the coming months.
Why it matters
A chatbot that can produce fluent Azerbaijani is not automatically one that understands the needs of people using public services. Hasanov’s example was advice about caring for an elderly parent: a generic answer might recommend professional care, while overlooking that family care is the norm in his society. The point is not that one answer fits every Azerbaijani family, but that language and cultural assumptions can shape what a system presents as sensible advice.
Building local-language capability could make government tools more accessible and give the state more control over how they work. It also raises a harder question: how will officials show that these models provide accurate, useful service rather than simply reflecting preferred national values?
Our read
This is more than a chatbot plan. Azerbaijan is linking local-language AI to public-service access, existing tools on my.gov and civil-service training. That is a meaningful direction, though the announcement supplies no model names, performance results or deployment timetable for the new systems. The proof will be in whether people can get better answers in Azerbaijani, not in how confidently a strategy document says “sovereignty”.
What to watch
- Which Azerbaijani-language models are built, and when they will be used in state services.
- What new AI services arrive on my.gov by year-end, and how people can access them.
- Whether the government publishes evidence about answer quality across languages and user groups.
- How the planned training for 25,000 civil servants is delivered and assessed.
Discussion spark: Should governments build their own language models for public services, even if global systems may be more capable, or should they focus on testing and improving whichever models work best?
Sources and evidence
- Azerbaijan prepares native-language AI models for state services – Trend.Az (27 October 2026, 09:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.