Latvia - AI Regulation Overview

Overview

Latvia’s journey toward a comprehensive artificial intelligence regulatory framework is a testament to its ambition to transition from a digital-first society to an AI-integrated economy. The evolution began in earnest with the 2020 Informative Report 'On the Development of Artificial Intelligence Solutions,' which laid the conceptual groundwork for how the state should approach automated decision-making. Since then, the Latvian government, led by the Ministry of Smart Administration and Regional Development (VARAM), has moved with significant speed to codify these principles into binding law. The centerpiece of this effort is the Law on the Latvian Artificial Intelligence Centre, adopted in 2025, which represents a shift from passive observation of technological trends to active, institutionalized management of AI risks and opportunities. This framework is not merely a reaction to technological change but a proactive attempt to define the 'Latvian model' of AI governance—one that is deeply integrated with European values, focused on linguistic and cultural preservation, and uncompromising on the integrity of democratic institutions. By establishing a dedicated national center, Latvia has signaled its intent to provide a stable and predictable legal environment for both public sector innovators and private enterprises, ensuring that the deployment of AI systems is consistent with the rule of law and the protection of individual liberties.

Regulatory Approach

The Latvian regulatory philosophy is defined by a 'hybrid-horizontal' model. This means that while there is a central, horizontal framework—the Law on the Development of Artificial Intelligence—that applies across all sectors, it is supplemented by highly specific, risk-based interventions in areas where AI poses the greatest threat to social stability. This approach avoids the pitfalls of a fragmented, purely sectoral regime while providing the necessary granularity for high-stakes applications. A defining feature of this approach is the statutory creation of 'special regulatory environments,' or sandboxes. These are not merely informal testing grounds but legally defined spaces where the Latvian Artificial Intelligence Centre can authorize the testing of innovative AI systems under controlled conditions. Within these sandboxes, certain regulatory requirements may be temporarily relaxed to allow for real-world validation, provided that strict safety protocols and data protection measures are maintained. This 'innovation-first' mindset is balanced by a 'safety-always' requirement, where the burden of proof for the safety of a high-risk system remains with the developer. Furthermore, the Latvian approach is characterized by a high degree of institutional coordination, where the AI Centre acts as a bridge between the technical expertise of academia and the regulatory authority of the state, ensuring that policy is informed by the latest scientific developments.

Key AI Legislation

The legislative landscape in Latvia is anchored by several key instruments that work in concert to regulate the AI lifecycle. The Law on the Latvian Artificial Intelligence Centre (2025) is the foundational act, establishing the Centre as a public foundation with the mandate to coordinate national AI policy, manage the regulatory sandbox, and provide expert opinions to other state agencies. This is complemented by the Law on the Development of Artificial Intelligence (2025), which sets out the broader strategic objectives for the national AI ecosystem, including the promotion of AI literacy and the ethical use of automated systems in public administration. In response to the specific threat of disinformation, the 2024 Amendments to the Pre-election Agitation Law introduced mandatory disclosure requirements for AI-generated content in political advertising, ensuring that voters are aware when they are interacting with synthetic media. Simultaneously, the 2024 Amendments to the Criminal Law (Article 90.1) provided the necessary 'teeth' to this framework by criminalizing the malicious use of deepfakes to influence elections. Older but still relevant documents, such as the Digital Transformation Guidelines 2021–2027 and the 2020 Informative Report, provide the policy context and long-term vision that continue to guide the interpretation and implementation of these newer, binding statutes. Together, these laws form a comprehensive net that captures both the developmental potential and the inherent risks of AI technology.

Governance & Enforcement Bodies

Governance in Latvia is a multi-tiered affair, designed to ensure that no single agency has an unchecked monopoly on AI oversight. At the strategic level, the Ministry of Smart Administration and Regional Development (VARAM) remains the primary architect of digital policy, responsible for the overall implementation of the National AI Strategy and the allocation of state resources. However, the operational heart of the system is the Latvian Artificial Intelligence Centre. This Centre is governed by a Supervisory Board that brings together the diverse interests of the Ministry of Economics, the Ministry of Defense, and the Ministry of Smart Administration, alongside independent experts. This structure ensures that AI policy is not siloed but reflects the cross-cutting nature of the technology. The State Digital Development Agency (VRAA) provides the critical 'back-office' support, managing the IT infrastructure and ensuring that the Centre's operations meet the highest standards of cybersecurity and administrative efficiency. For specialized enforcement, the Corruption Prevention and Combating Bureau (KNAB) monitors the use of AI in political campaigns, while the Data State Inspectorate (DVI) serves as the ultimate arbiter of privacy and data protection. This distributed model leverages existing institutional strengths while centralizing technical AI expertise within the AI Centre, creating a robust and resilient governance network.

Penalties & Enforcement

Latvia has adopted a tiered enforcement strategy that distinguishes between administrative non-compliance and criminal misconduct. The most severe penalties are reserved for actions that threaten the democratic foundations of the state. Under the amended Criminal Law, the use of deepfake technology to intentionally mislead the public or discredit candidates during an election period is a serious felony, punishable by up to five years of imprisonment. This high penalty reflects the state's view of AI-enabled disinformation as a threat to national security. On the administrative side, the Corruption Prevention and Combating Bureau (KNAB) has the power to issue immediate 'cease and desist' orders for non-compliant AI content, with failure to comply leading to significant fines. The Data State Inspectorate (DVI) also wields substantial power, with the ability to impose fines under the GDPR framework that can reach millions of euros or a percentage of a company's global turnover. Furthermore, the Law on the AI Centre introduces a unique enforcement mechanism for the regulatory sandbox: if a participant violates the terms of the 'special regulatory environment,' the Centre can not only terminate the project but also demand the repayment of any state subsidies and bar the entity from future participation. This combination of criminal, administrative, and contractual penalties ensures that all actors in the AI ecosystem have a strong incentive to adhere to ethical and legal standards.

Data Protection Framework

Data is the lifeblood of AI, and Latvia's framework for data protection is designed to be both rigorous and enabling. As an EU member, Latvia's primary data law is the GDPR, which is supplemented by the national Personal Data Processing Law. The Latvian approach emphasizes the principle of 'purpose limitation' and 'data minimization,' particularly when public sector data is used to train AI models. The Law on the Latvian Artificial Intelligence Centre introduces specific safeguards for the processing of data within the national AI sandbox, requiring that all personal data be pseudonymized or anonymized by default. A key pillar of the future data landscape is the National Data Governance Strategy, which envisions the creation of the 'DAGR' platform. This platform will serve as a secure, centralized gateway for the sharing of high-value datasets between government agencies and AI developers. By providing a clear legal and technical framework for data re-use, Latvia aims to overcome the data silos that often hinder AI innovation. The Data State Inspectorate (DVI) plays a proactive role in this ecosystem, providing guidance on how to conduct Data Protection Impact Assessments (DPIAs) for AI systems and ensuring that the rights of data subjects—including the right to an explanation for automated decisions—are fully respected in the age of machine learning.

Sector-Specific Rules

While Latvia maintains a horizontal oversight model, it has pioneered sector-specific rules in areas of critical national interest. The most developed of these is the regulation of the political and electoral sector. The Pre-election Agitation Law now includes specific provisions that require any AI-generated audio, visual, or text content used in paid advertisements to be accompanied by a clear and prominent disclaimer. This is intended to protect the 'cognitive sovereignty' of the Latvian electorate. In the public sector, there are specific guidelines for the use of AI in citizen services, such as the 'UNA' virtual assistant, which must comply with strict transparency and accessibility standards. Another unique sectoral focus is the preservation of the Latvian language. The AI Centre is specifically tasked with supporting the development of Large Language Models (LLMs) and other NLP tools that are optimized for the Latvian language, ensuring that the country's linguistic heritage is not marginalized by global AI platforms. While sectors like healthcare and finance are currently governed by broader EU regulations, the Latvian framework allows for the creation of 'sectoral annexes' to the national AI strategy, which will provide more tailored guidance as these technologies become more prevalent in clinical diagnostics and automated financial services.

International Alignment

Latvia's AI strategy is inextricably linked to its membership in the European Union and its commitment to international norms. The 2025 AI legislation was specifically designed to be 'future-proofed' against the EU AI Act, ensuring that Latvia's national authorities and enforcement mechanisms are ready to fulfill their roles as soon as the EU-wide regulations become fully applicable. Latvia is an active participant in the EU's Coordinated Plan on Artificial Intelligence and frequently collaborates with its Baltic and Nordic neighbors on regional AI initiatives. Beyond the EU, Latvia adheres to the OECD Principles on Artificial Intelligence, which emphasize human-centric values, transparency, and accountability. The country also views AI through the lens of international security, participating in NATO-led discussions on the ethical use of autonomous systems and the defense against AI-driven hybrid threats. This international alignment is not just about compliance; it is a strategic choice to ensure that Latvian AI companies can easily access the European Single Market and that the country remains an attractive destination for international research and investment. By adopting global standards for 'Trustworthy AI,' Latvia positions itself as a reliable partner in the global effort to govern emerging technologies.

Future Developments

The next two years will be a period of intense implementation for Latvia's AI framework. By mid-2025, the Cabinet of Ministers is expected to finalize a series of detailed regulations that will breathe life into the Law on the AI Centre. These will include the specific technical requirements for the regulatory sandbox, the criteria for selecting 'high-impact' AI projects for state support, and the procedures for inter-agency data sharing. The Latvian Artificial Intelligence Centre is expected to reach full operational capacity by 2026, at which point it will begin issuing formal certifications for AI systems and conducting regular audits of high-risk applications in the public sector. There is also an ongoing legislative debate regarding the need for a dedicated 'AI Liability Act' to clarify who is responsible when an autonomous system causes harm, particularly in the context of healthcare or transportation. Furthermore, as the EU AI Act's requirements for 'General Purpose AI' (GPAI) models come into effect, Latvia will likely refine its national monitoring capabilities to ensure that large-scale models deployed within its borders comply with European transparency standards. These upcoming developments suggest that Latvia's AI law is a living framework, designed to evolve in tandem with the technology it seeks to govern.

Key Regulations

Enforcement Bodies