Kazakhstan’s AI Law: What It Will Change

On November 17, 2025, President Kassym-Jomart Tokayev signed the Law “On Artificial Intelligence.” The law will come into force on January 18.

“Artificial intelligence systems are defined as objects of informatization. From now on, AI will be recognized as a tool for achieving specific human tasks,” the Akorda stated.

The law, as declared, enshrines principles of responsibility and accountability, as well as legality, fairness, equality, transparency and explainability, prioritization of human well-being, freedom of choice in decision-making, data protection and confidentiality, safety and security. It consists of a total of 31 articles.

Factcheck.kz examines what this law will change and how it will affect the digital life of Kazakhstanis.

Labeling — Mandatory

Perhaps one of the most significant provisions for internet users and content creators using AI is Article 21 — “Ensuring Transparency in the Use of Artificial Intelligence Systems.” According to it, users must be informed that goods, works, and services are produced or provided using AI systems.

Clause 2 of the same article states:

“The dissemination of synthetic results of AI systems is permitted only if they are labeled in machine-readable form and accompanied by a visual or other warning that allows users to perceive them without methods that hinder such perception.”

Clause 3 places responsibility for compliance with these requirements on AI system owners.

From the perspective of combating disinformation, this provision is useful. It introduces rules aimed at protecting users from manipulation and the spread of false information, including deepfakes and automatically generated content.

Machine-readable labeling allows:

• easier detection of AI content by platforms and fact-checkers;

• tracking the spread of synthetic materials;

• reducing the risk of mass distribution of deepfakes in media and social networks.

Copyright and AI

Article 23 addresses one of the most controversial issues in the AI debate — who owns copyright for content created with AI. According to Clause 1:

Works created using AI systems are protected by copyright only if there is human creative input in their creation.

Clause 2 recognizes text prompts as copyrightable objects if they are the result of human intellectual creative activity.

However, both points raise questions. First, how can a person prove creative input in a work created with AI? Second, how can an author know and prove that someone else appropriated their prompt?

Clause 3 excludes the use of works for AI training from the list of cases of free use for educational and scientific purposes.

Clause 4 states that using works (texts, images, music, etc.) for AI training does not fall under the usual copyright actions: reproduction, distribution, adaptation, public display, etc.

Finally, Clause 5 notes: “The use of works for training AI models is permitted only if there is no prohibition from the author or rights holder, expressed in machine-readable form.” This may be problematic, since unlike large platforms and media, many individual authors publishing works online (illustrations, photos, small works) may not be able to express such a prohibition in machine-readable form.

Other User Rights

Article 16 regulates the basic rules of human interaction with AI systems. It enshrines the right to understand what exactly, under what conditions, and with what consequences a person interacts with. Users have the right to know the rules of system functioning and demand protection of their personal data and confidential information.

Users also have the right to receive explanations for decisions made using AI if they affect their rights and legitimate interests. This is especially important when algorithms are involved in evaluation, selection, or decision-making — for example, in services, finance, content, or access to information.

Users also have the right to request information about the data on which AI made a decision. An important element of the article is the right to refuse interaction with AI (if such interaction is not mandatory).

Advanced, but Not Without Problems

Yelzhan Kabyshev, head of legal practice at the Eurasian Digital Foundation, who participated in the working group on drafting the law, says that the organization’s key recommendations were included in the final text.

According to him, Kazakhstan’s model shows some harmonization with advanced practices, but with national specifics. He notes that Kazakhstan adapted the risk-based approach from the EU AI Act: classification of systems by risk levels and a “blacklist” of prohibited practices (manipulation, social scoring) are almost identical to the European model. But there is a fundamental difference.

Compliance with OECD and UNESCO principles is ensured through transparency, accountability, and human-centricity enshrined in the law. The main difference lies in the institutional part: instead of independent supervisory bodies as in the EU, Kazakhstan uses centralized management through a specialized Ministry and Platform Operator, apparently due to the need for accelerated digital development.

Kabyshev notes that the law builds a model of state-infrastructure partnership. The balance between state, business, and human rights is achieved through compromise: the state acts not only as regulator but also as market driver, creating a National AI Platform and providing businesses with computing power (a supercomputer), which hypothetically lowers entry barriers for startups.

Business gains legal certainty and resources but must comply with legislation, especially labeling of generated content and safety requirements. Citizens’ rights are protected through bans and transparency requirements, though the state’s dual role as platform operator and regulator makes it the dominant player.

At the same time, Kabyshev raises concerns about biometric data privacy.

“We positively assess the direct legislative ban on manipulative techniques and social scoring. However, the practice of emotion recognition (Article 17) remains an area of monitoring. The law allows this technology with consent. Our task as experts is to ensure that in workplaces, schools, and other places this consent is truly voluntary, not forced, and substantively justified,” he said.

He explains that EDF experts insisted on a full ban on biometric identification in public places, except for strictly defined security purposes. Lawmakers partially accepted this: use of such systems is limited to “protection of the constitutional order and law enforcement.”

By manipulative techniques, Kabyshev refers to Article 17, which prohibits using AI for hidden influence on human subconsciousness or exploiting vulnerabilities.

What Punishment Is Provided

The Code of Administrative Offenses was supplemented with Article 641-1 “Violation of the Legislation of the Republic of Kazakhstan in the Field of Artificial Intelligence.”

According to it, failure by AI system owners to inform users about the synthetic nature of content, or failure to fulfill obligations to manage risks of high-risk AI systems that negatively affect health or well-being, create or spread prohibited or false information, discrimination, or human rights violations, if not criminally punishable, entails fines: 15 monthly calculation indices (MCI) for individuals and officials; 20 for small businesses or non-profits; 30 for medium businesses; 100 for large businesses.

Repeated violation within a year after administrative penalty entails fines: 30 MCI for individuals; 50 for small businesses or non-profits; 70 for medium businesses; 200 for large businesses, with suspension or prohibition of the AI system’s activity.

Conclusion

Against the backdrop of rapid AI development in recent years, governments worldwide are increasingly adopting laws, policies, and strategies to regulate this new technology. According to OECD data, 72 countries and territories currently have specific AI policies. Kazakhstan is among the states that adopted a separate law. However, given the absence of a universal global model and the ongoing evolution of AI, national legislation in this area will likely be revised and supplemented multiple times.

Source: Factcheck.kz — https://factcheck.kz/prava-cheloveka/zakon-ob-ii-v-kazahstane-chto-on-izmenit/