Country Profile – Estonia

Estonia has earned international recognition for its ambitious and comprehensive e-governance ecosystem, including as a globally recognized pioneer in the digitization of democratic processes. With the introduction of the landmark i-voting system in 2005, Estonia became the first country in the world to allow constituents to cast their votes in a digital ballot box. After a cautious initial uptake, digital voting has expanded steadily and, by the 2023 parliamentary election, had become the most popular voting method among constituents (Valimised Eestis ‘Statistics’ 2026). Building on this experience, Estonia continues to assess how emerging technologies can enhance democratic accessibility and administrative efficiency without compromising the integrity of electoral functions. 
In recent years, artificial intelligence (AI) has featured more prominently in these discussions within the Estonian government. Yet many questions are still under consideration, regarding the safety, utility and legal implications of artificial intelligence. Estonia’s approach to AI in electoral management reflects the holistic philosophy that underpins the country’s broader digital transformation. This model is based on a synergistic framework for developing and governing digital public services (Ministry of Economic Affairs and Communications 2024a; Justiits- ja Digiministeerium n.d.). 

Who governs electoral AI in Estonia?

Under the Estonian Election Act 2024, the State Electoral Office (SEO; Valimised Eestis) is charged with organizing and supervising the electronic voting process (Riigikogu 2024). Together with the Republic of Estonia’s Information System Authority (Riigi Infosüsteemi Amet; RIA), which manages the technical development of election technologies, the SEO has developed an intricate digital architecture to ensure the protection of sensitive electoral information, from ballots to voter data (Information System Authority 2025). 

The SEO’s mandate to safeguard the integrity of electoral information has given way to a prudent approach to AI adoption, stressing the need to engineer adequate data protection and validation standards before introducing AI in any substantive applications. At present, no AI systems are deployed in core election administration or voting processes. However, initial talks on the role of AI in elections are taking place at the parliamentary level, particularly on issues of personal data sovereignty and management (Majandus-ja Kommunikatsiooniministeerium 2024b). 

In 2024, the Estonian Ministry of Economic Affairs and Communications published a pan-governmental white paper on Artificial Intelligence and data for the years 2024–2030. The strategy introduces a collaborative approach to AI among public bodies, positioning inter-agency cooperation in technology at the forefront of developing, regulating, implementing and overseeing AI in the public sector. This illustrates an approach in which openness to AI innovation is matched by institutional restraint, positioning the potential of electoral AI as contingent on robust governance and the mitigation of risk. 

How is AI used to improve electoral management in Estonia?

The SEO has limited its use of AI to internal, non-decision-making functions aimed at supporting staff capacity. Currently, the SEO is utilizing AI in two main ways:

• As productivity tools: Institution-wide deployment of productivity tools such as Microsoft Copilot assists with notetaking, document editing and internal communication. These tools are intended to streamline routine tasks rather than substitute for professional judgement.
• As legal research and public inquiry response: The SEO licences the commercial legal research platform ExtendLaw (xLaw) to support the preliminary analysis of complaints submitted to the Estonian National Election Commission (NEC). ExtendLaw functions as a plug-in search engine, enhanced by large language model (LLM) capabilities, for retrieving relevant provisions of Estonian electoral law and Supreme Court jurisprudence when prompted with user queries. The main purpose of the tool is to support the NEC during the initial steps of complaints processing, but the final response given to the inquirer is always tailored by legal experts. Since the numbers of inquiries typically surge around the time of elections, the SEO uses AI tools to smooth out workload pressures and support staff capacity.

The SEO emphasizes that expert human oversight is a crucial intermediate failsafe for current AI applications, given the known limitations of LLMs—including plausible but misleading outputs. This is particularly important when applied to electoral management, as AI misconstructions— ‘hallucinations’—risk undermining the fairness of democratic procedures or failures in explaining public decision making.

Areas where AI tools are currently under consideration

Given the high-risk nature of AI, there are currently no plans to integrate it into Estonia’s voting arrangements. Unresolved concerns around data security and sensitivity remain decisive constraints. However, the SEO acknowledges that there are several potential areas where the integration of AI systems may be beneficial, particularly public communications and voter information services:

• Voter information: The SEO is currently reviewing the use of AI tools as additional pathways for constituents to access trustworthy, reliable and verified information about Estonian elections. Crucially, such systems would operate alongside existing call centres, email channels and website-integrated AI chatbots, improving the response times for routine questions on matters such as election dates or polling locations.
• Public inquiry triage: In addition to offering additional communication channels for voters, AI agents could support SEO staff by screening incoming requests to help interpret the meaning of unclear questions, or, via an AI call-centre triage system, rerouting callers to the appropriate department. 

The SEO stresses that AI-agent systems must coexist with traditional channels rather than replace them. All AI tools require automatic mechanisms that redirect users to a human contact point for complex questions if the user expresses displeasure with their AI experience. This ensures not only that there are roll-back mechanisms in case of system malfunction, but also that public trust is maintained by offering opt-out alternatives for voters who are skeptical, uncomfortable or unaccustomed to AI interactions. 

While the SEO acknowledges that AI systems can contribute initial filters, consultation or clarification, they must ultimately serve the real needs and demands of the parties at each end of the line of communication—the election official and the voter. In other words, model outputs must be factually correct, users must be interested in engaging with non-human systems and responses must contain information relevant to the question asked.

Inter-agency collaboration as an approach to electoral AI

The SEO relies on a collaborative and modular approach to election technologies, leveraging partnerships with other state agencies rather than maintaining an in-house IT unit. While the SEO is responsible for supervising the i-voting systems and other election technologies, the development and technical implementation is conducted by the RIA. These partnerships constitute the foundation for any election-related AI projects, rooted in broader government-led initiatives. 

This interconnected approach encompasses the SEO’s work on raising institutional awareness of, and literacy about, AI. Staff representatives have received training on how to use licensed AI products together with the the Chancellery of the Riigikogu, the Estonian parliament (Inter-Parliamentary Union 2025), of which the SEO is an independent structural unit (Valimised Eestis ‘State Electoral Office’). Following this training, the Chancellery has made initial commitments to develop further internal recommendations that will take inspiration from parliamentary guidelines on AI in other countries and the Resolution on ‘The impact of artificial intelligence on democracy, human rights and the rule of law’ adopted by the Inter-Parliamentary Union in October 2024 (Inter-Parliamentary Union 2024 and 2025). Such future parliamentary guidelines will be decisive in determining the role of AI in electoral management at the SEO.

Response to AI use by other electoral actors

Estonia’s holistic model of digital governance is further evidenced by its response to the use of AI by other actors in the electoral information environment. The threat of Russian interference in elections had a foundational influence on Estonia’s early digitization efforts. In 2007, a series of globally unprecedented DDoS (distributed denial-of-service) attacks, coordinated by Russian actors, prompted a redoubled focus on Estonia’s cyber resilience (StratCom n.d.). 

Today, however, the reported cases of AI-enabled disruption have been anecdotal, and no centrally orchestrated campaigns have been found to compromise the integrity of recent elections (Estonian Academy of Sciences, 2024). The type of disruptive AI-generated content that has gained some visibility online consists primarily of so-called ‘foefakes’—clearly exaggerated AI-generated images depicting political figures in a mocking or satirical manner, intended to influence public sentiment rather than to mislead audiences in a credible way (Wack et al. 2025). 

Estonian state institutions have not experienced large-scale coordinated attacks, and the impact of AI-driven disinformation on the public’s trust in democratic institutions has remained limited. As a result, the SEO has not issued any formal statements regarding the influence of AI-generated content in elections. Nevertheless, the potential of AI-enabled threats to electoral integrity remains a significant concern, especially as Estonia looks ahead to the 2027 parliamentary elections. In the lead-up to the last municipal elections in October 2025, the Chancellery of the Riigikogu engaged in discussions with several major platform owners to explore strategies for mitigating the risks of electoral misinformation. These conversations have carried on, as Estonia continues to refine its approach to AI-driven disruptions in the electoral process.

Ethical concerns, data protection and cybersecurity

For the SEO, the protection of voters’ personal data is a fundamental concern in the consideration and appraisal of any AI tool. Under the Riigikogu Election Act 2024, Estonia’s electronic voting system adheres to strict principles of data minimization and anonymization (Riigikogu  2024). All personal data collected during elections, by both the SEO and any other parties, are destroyed following completion of the electoral process. Consequently, AI tools reliant on access to personal data, including in development and training, risk conflicting with the specific statutory obligations of Estonia’s election law unless data are fully encrypted.

Voter data are inherently a high-value target for cyber-attacks because of their sensitivity. The data sets used for training AI systems could become vulnerable to data brokers or malicious actors. The 2024 white paper on AI and data (Majandus-ja Kommunikatsiooniministeerium 2024b) showcases Estonia’s acute awareness of the risks associated with the governance of citizens’ data, placing focus on human-centric data collection, informed consent, accessibility, and stringent data protection standards. The SEO continues to highlight the significant legal and ethical risks associated with utilising voter data for training AI models. All legal conflicts must be resolved unequivocally—including compliance with the General Data Protection Regulation (GDPR) by way of Estonia’s Personal Data Protection Act 2018, limitations around intended purpose, the retention of data and obligations to delete data—before the SEO will consider developing or piloting an AI tool that involves personal data. Opening up to AI tools is not merely a question of technical capacity, literacy and awareness, nor one of public demand; rather, it is a question of regulatory compliance and the protection of democratic principles.