Published
09/28/2026, 14:22Is the person replying to you in the chat a human or a programme? Is the photograph you’re looking at genuine, or was it generated by a neural network? In Kyrgyzstan, there is a proposal to make it compulsory for owners of AI systems to disclose this information. The draft legislation was drawn up by the Ministry of Digital Development, which no longer exists, having been abolished by a presidential decree dated 30 April this year. Its functions and responsibilities have been transferred to the Office of the President.
The public consultation will run until 13 October 2026.
Under the draft, text, photographs, videos and audio recordings that have been entirely created or substantially altered using AI will need to be labelled. This could take the form of a visible label, an audio warning or information in the metadata — the technical details of the file.
Chatbots, voice assistants and automated calling systems will be required to inform users in advance that they are interacting with a programme. Such notification must be provided before operations commence or data is collected.
A separate proposal calls for a ban on concealing the use of AI if the content it generates could mislead the public, infringe citizens’ rights, influence elections or harm national security. Exceptions to the labelling and notification requirements are permitted for scientific research, technology trials under special legal regimes, and other cases provided for by law.
People and organisations will be held liable for errors made by artificial intelligence. The draft legislation establishes the liability of developers, owners and users, taking into account their respective roles in the system’s operation. AI itself is not recognised as a legal entity, which means that liability for any damage caused cannot be shifted onto the programme.
State and local authorities will be required to use systems that have undergone ethical and safety assessments. Such assessments will also become mandatory for high-risk systems. For other private businesses, it is recommended that they carry out these assessments on a voluntary basis.
The AI will be assessed by the developers, suppliers or owners themselves — prior to launch and following any significant update. They will need to check whether the system could violate human rights or give rise to discrimination, and to what extent its decisions remain under human control.
The results must be recorded in a special document – an AI ethics declaration. Publication is not mandatory, but in cases specified by law, the document must be provided upon request by authorised state bodies.
Another requirement concerns the Kyrgyz language. AI services for state and municipal services, public information, education, healthcare and public infrastructure must function fully in Kyrgyz on a par with Russian. Developers are also advised to improve the quality of the Kyrgyz-language materials used to train neural networks.
A separate section is devoted to history and culture. The draft prohibits the creation of knowingly false information that defames historical figures, state and cultural figures of Kyrgyzstan, or belittles their historical achievements. To this end, developers are encouraged to implement technical restrictions and filters. The use of images of historical figures is permitted for scientific, educational, museum, documentary and public awareness purposes — provided that historical accuracy is maintained.
Owners of AI systems must also protect them against hacking and interference, and comply with legislation when handling personal data. The document recommends anonymising data and regularly checking neural networks for errors and ‘hallucinations’ – instances where the programme presents fabricated information as fact.
The proposed rules concern the civilian use of AI. The provisions will not apply to defence, national security, law enforcement or operational and investigative activities.
According to the Ministry’s assessment, the draft will not require any additional expenditure from the national budget. The drafter also considers that the document is not intended to regulate business, and therefore no analysis of its impact on business has been carried out. However, some of the requirements do affect private companies working with AI.
