Kazakhstan - Ethical AI Guidelines

Overview

The National Ethical Code in the Field of Artificial Intelligence (the Code) represents Kazakhstan's primary soft-law instrument designed to guide the responsible development, deployment, and use of artificial intelligence technologies. Formally introduced in 2024 by the Ministry of Digital Development, Innovations and Aerospace Industry (MDDIAI), the Code serves as a moral and professional foundation for the country's rapidly expanding digital ecosystem. It was developed in response to the directives of President Kassym-Jomart Tokayev, who emphasized the need for a balanced approach that promotes technological leadership in Central Asia while safeguarding the fundamental rights of Kazakhstani citizens. The Code is a voluntary framework intended for adoption by government agencies, private enterprises, research institutions, and individual developers. As Kazakhstan transitions toward a more AI-integrated economy, the Code acts as a bridge between the general Law on Informatisation (2015) and the more prescriptive Law on Artificial Intelligence (2025). By establishing a shared set of values, the Code aims to foster public trust in automated systems and mitigate risks associated with algorithmic bias, privacy violations, and lack of accountability. It aligns with the 'Concept for the Development of Artificial Intelligence for 2024–2029,' positioning ethics as a prerequisite for innovation rather than a barrier to it. The document is structured around six core principles that harmonize national cultural values with international standards, such as those proposed by UNESCO and the OECD.

Definitions

The Code provides a conceptual framework for understanding the ethical dimensions of AI within the Kazakhstani legal context. It defines an 'Artificial Intelligence System' not merely as a technical tool, but as a socio-technical construct that can impact human behavior and decision-making. Central to the document is the definition of 'Ethical AI,' which is described as the development and application of technologies in a manner that respects human dignity, ensures social justice, and promotes the collective good. Unlike technical standards, these definitions focus on the 'intent' and 'impact' of the technology rather than its underlying code or architecture. Furthermore, the Code introduces the concept of 'Stakeholders' in the AI lifecycle, which includes developers, operators, and end-users. It emphasizes the 'Human-in-the-Loop' (HITL) requirement, defining it as a mandatory oversight mechanism where humans retain the final authority over life-altering decisions made by AI. Another key definition is 'Algorithmic Fairness,' which refers to the active prevention of discriminatory outcomes based on race, gender, religion, or socio-economic status. These definitions are intended to be interpreted broadly to remain relevant as AI capabilities evolve from simple machine learning models to more complex generative and autonomous systems, ensuring that the legal and ethical vocabulary keeps pace with rapid technological shifts.

Governance and Institutional Framework

The governance of AI ethics in Kazakhstan is centralized under the Ministry of Digital Development, Innovations and Aerospace Industry (MDDIAI). Within the Ministry, the 'AI and Innovation Development Committee' (established in May 2024) is the primary body responsible for the implementation and promotion of the Code. This committee coordinates with other state bodies to ensure that ethical considerations are integrated into public procurement and the delivery of digital government services. The institutional framework also includes an 'Expert Council on AI,' composed of scientists, legal experts, and industry representatives, who provide recommendations on complex ethical dilemmas and the evolution of the Code. A unique feature of the Kazakhstani framework is the role of the 'Ethics Commissioner' within organizations that adopt the Code. These individuals act as internal ombudsmen, ensuring that AI projects undergo ethical impact assessments before deployment. The governance model favors a 'self-regulatory' approach, where the government provides the standards and the private sector demonstrates compliance through voluntary certification and public declarations. This 'soft' governance is intended to be agile, allowing the MDDIAI to update guidelines as new risks emerge, such as those posed by deepfakes or large language models, without the lengthy legislative process required for formal laws. This structure ensures that the state maintains oversight while allowing the market to innovate within a safe and predictable ethical perimeter.

The Six Core Ethical Principles

The Code is built upon six foundational principles that serve as the 'moral compass' for AI development in Kazakhstan. The first is Human-Centricity, which mandates that AI must serve the interests of people and protect their constitutional rights. The second is Safety and Reliability, requiring that systems be technically robust and include fail-safes to prevent physical or psychological harm. The third principle is Transparency and Explainability, ensuring that AI operations are not 'black boxes' and that users are informed when they interact with automated systems. The fourth principle is Accountability, which establishes that humans are always responsible for the outcomes of AI systems, regardless of the complexity of the algorithm. The fifth is Fairness and Non-Discrimination, which requires developers to actively identify and mitigate biases in training data that could lead to unequal treatment of different social groups. Finally, the principle of Environmental Sustainability encourages the development of energy-efficient AI models to minimize the ecological footprint of large-scale data processing. These principles are not merely theoretical; they are intended to be integrated into the technical specifications of every AI project initiated within the country, creating a culture of 'Ethics by Design' that permeates the entire technological lifecycle from data collection to decommissioning.

Key Focus Areas and Prohibitions

The Code identifies several critical areas where ethical vigilance is paramount. The first is 'Human Rights and Freedoms,' which mandates that AI must never be used to undermine the constitutional rights of citizens. This includes a strict prohibition on the use of AI for mass surveillance that violates privacy or for social scoring systems that could lead to systemic marginalization. The Code places a high priority on 'Safety and Reliability,' requiring that AI systems be tested for robustness against adversarial attacks and that they have clear 'kill-switch' mechanisms to prevent unintended harm in critical sectors like healthcare or transportation. Another major focus is 'Transparency and Explainability.' The Code stipulates that users must be clearly informed when they are interacting with an AI system (e.g., chatbots or synthetic media). For AI-driven decisions that affect an individual's legal or financial status, the system must be capable of providing a human-readable explanation of the logic used. Additionally, the Code addresses 'Data Sovereignty and Privacy,' emphasizing that the training of AI models must respect national data protection laws, ensuring that personal data is processed only with informed consent and for specified, legitimate purposes. These focus areas are designed to address the most pressing public concerns regarding the rapid adoption of AI in public services and the private sector.

Implementation Framework and Signatory Model

The implementation of the National Ethical Code is primarily voluntary, relying on a 'Signatory Model.' Organizations—ranging from tech giants to startups—are encouraged to formally sign the Code at major national forums like the 'Digital Bridge' or 'AI Journey.' By signing, these entities commit to integrating the Code’s principles into their internal corporate policies, software development life cycles (SDLC), and employee training programs. The MDDIAI provides a toolkit for implementation, which includes templates for ethical impact assessments and guidelines for establishing internal ethics committees. To incentivize adoption, the government of Kazakhstan is exploring the integration of ethical compliance into its 'trusted software' registry. Companies that can demonstrate adherence to the Code may receive preferential treatment in government tenders or access to state-funded AI infrastructure, such as the National AI Platform and the national supercomputer. This 'comply-or-explain' approach encourages a culture of responsibility within the tech community. Furthermore, the implementation framework encourages the development of 'Ethics by Design,' where considerations of bias and safety are baked into the AI model from the initial data collection phase through to final deployment. This proactive approach aims to reduce the long-term costs of compliance by preventing ethical failures before they occur.

Monitoring, Evaluation, and Public Oversight

Monitoring the effectiveness of the Code is a multi-stakeholder process involving the government, civil society, and the private sector. The AI and Innovation Development Committee conducts periodic reviews of how the Code is being applied in practice. This includes analyzing reports from Ethics Commissioners and monitoring public sentiment regarding AI deployments. While there is no formal audit process for this voluntary code, the government maintains a public list of signatories to ensure transparency and allow for public oversight. Civil society organizations are encouraged to act as 'watchdogs,' reporting instances where signatories may have violated the spirit of the Code. Evaluation also involves measuring the impact of the Code on the national AI ecosystem. The MDDIAI tracks metrics such as the number of organizations that have adopted the Code, the frequency of AI-related ethical complaints, and the degree to which Kazakhstani AI products meet international ethical benchmarks. These evaluations inform the periodic updates to the Code, ensuring it remains a 'living document.' The goal is to create a feedback loop where practical challenges in implementation lead to clearer guidelines and, eventually, more robust legal protections in the formal Law on AI. This iterative process allows the state to refine its regulatory approach based on real-world data and stakeholder feedback.

Penalties, Liability, and Redress Mechanisms

As a 'soft law' instrument, the National Ethical Code does not carry direct administrative or criminal penalties for non-compliance. Instead, the primary consequences for violating the Code are reputational and commercial. Signatories who fail to uphold the principles risk being removed from the official list of ethical AI providers, which can lead to a loss of public trust and exclusion from certain government-sponsored innovation programs. However, it is important to note that many of the ethical principles in the Code (such as data protection and non-discrimination) are also enshrined in binding laws, such as the Law on Personal Data and the Criminal Code, where violations can lead to significant fines or imprisonment. Regarding liability, the Code reinforces the principle that responsibility for the actions of an AI system always rests with a human or legal entity—never the AI itself. This 'human-centric' liability model ensures that victims of AI-driven harm have a clear path for redress. While the Code does not establish a specific appeals tribunal, it encourages organizations to create internal grievance mechanisms where users can challenge AI decisions. For more serious disputes, the general judicial system of Kazakhstan remains the final arbiter, with the Code serving as a reference for 'best practices' in determining whether a developer or operator acted with due diligence during the development and deployment phases.

Relationship to the Broader Legal Hierarchy

The National Ethical Code is a central component of a broader regulatory hierarchy in Kazakhstan. It operates alongside the 'Law on Informatisation' (2015), which provides the basic legal definitions for digital systems, and the 'Law on Personal Data and Their Protection.' Most importantly, the Code serves as the moral precursor to the 'Law on Artificial Intelligence' (2025). While the 2025 Law provides the 'hard' rules—such as mandatory risk assessments for high-risk AI and penalties for unauthorized social scoring—the Code provides the broader ethical context and covers 'low-risk' applications where formal regulation might be too burdensome. The Code also complements the 'Concept for the Development of Artificial Intelligence for 2024–2029,' which outlines the country's strategic goals for infrastructure, human capital, and economic growth. By aligning these documents, Kazakhstan ensures that its push for technological advancement is not decoupled from its social and ethical values. The Code is also designed to be compatible with sectoral regulations, such as those emerging in the healthcare and financial sectors, providing a baseline of ethics that can be further refined by specific industry regulators. This layered approach allows for a flexible regulatory environment that can adapt to the specific needs of different industries while maintaining a consistent ethical standard across the entire national economy.

International Alignment and Global Cooperation

Kazakhstan has explicitly designed the National Ethical Code to align with global standards to facilitate international cooperation and attract foreign investment. The document draws heavily from the 'UNESCO Recommendation on the Ethics of Artificial Intelligence' (2021), particularly regarding the protection of cultural diversity and the environment. It also incorporates the 'OECD AI Principles,' focusing on inclusive growth, human-centered values, and transparency. This alignment is intended to make Kazakhstan an attractive hub for international tech companies that require a stable and predictable ethical environment for their operations. Furthermore, the Code reflects Kazakhstan's ambitions to lead AI ethics discussions within regional organizations like the Shanghai Cooperation Organization (SCO) and the Organization of Turkic States. President Tokayev has even proposed the creation of an international commission on AI ethics to draft 'commandments for algorithms' at a global level. By adopting a Code that mirrors international best practices, Kazakhstan positions itself as a responsible actor in the global AI community, ensuring that its domestic AI products are 'ethical by design' and ready for export to markets with stringent regulatory requirements, such as the European Union. This strategic alignment not only protects domestic citizens but also enhances the global competitiveness of the Kazakhstani technology sector.

Implementation Timeline

Compliance Checklist