Bosnia's Shadow AI — A UNDP Survey Finds Most Civil Servants Already Use AI Tools With No Training, Prompting a Governed Rollout

In 2025, the United Nations Development Programme (UNDP) surveyed 516 civil servants across 57 state-level institutions in Bosnia and Herzegovina about their use of generative AI at work. The results, announced by Minister of Communications and Transport Edin Forto on 8 September 2025, showed that 69% of respondents already used AI tools — mostly free consumer services such as ChatGPT — in their daily work, entirely outside any government policy, licensing or security framework.
The same survey found that 95% of these civil servants had never received any formal training on using AI safely, even though 96% said they wanted it; asked what training they needed most, 75% pointed to practical tool use, 42% to basic AI concepts, and 40% to the legal and ethical rules around AI use. Independent Bosnian outlets covering the findings, including the investigative outlet Detektor (part of the BIRN network), warned that untrained staff feeding sensitive citizen and institutional data into foreign-hosted consumer AI tools created a real data-protection risk, given Bosnia and Herzegovina still has no national AI law or state-level AI strategy.
In response, the Ministry of Communications and Transport — described by Forto as moving the state administration from a "CIPS-paper era" to "verifiable and secure digital services" — is working with UNDP to license Microsoft's AI tools within a government-controlled environment and roll out structured training, positioning itself as a model other Bosnian state institutions could follow. As of the announcement, this remained an early-stage commitment: no institution-wide licensing numbers, training completion rates or measured productivity outcomes had yet been published.

Read the full analysis: https://detektor.ba/2025/10/16/vecina-drzavnih-sluzbenika-u-bih-koristi-umjetnu-inteligenciju-na-poslu-zasto-je-to-problem/?lang=en