Country Profile – Albania

In preparation for the upcoming 2027 local elections, the Albanian Central Election Commission (CEC) is actively scoping potential applications of electoral AI to assist in election management. In a recent public statement, the State Election Commissioner (SEC), Ilirjan Celibashi, noted that AI could support many aspects of the electoral administration, but he emphasized that its use must be measured against the specific requirements and risks associated with each practicable application (Si 2025). In line with this perspective, the CEC follows a rigorous system trial protocol before implementing any new AI technologies.

According to the Albanian Electoral Code, the CEC has the mandate to test and implement new electoral technologies. However, the full deployment of new technologies requires an initial pilot test covering 3 percent of the polling stations of any district (Albania 2008). If the results of the pilot evaluation are positive, the application’s general technical standards receive formal approval from the CEC’s Regulatory Commission (RC) and the SEC (CEC n.d.). These formal piloting and approval structures ensure that emerging AI procurement processes align with existing IT policy at the Commission and account for unforeseen risks of the kind that might accompany untested technologies.

How is AI used to improve electoral management in Albania?

The 2025 parliamentary and partial local (mayoral) elections marked the first time that the CEC undertook practical implementation of certain AI tools. The two key initiatives included a nationwide deployment of public-facing, AI-powered communication helplines, as well as an initial small-scale pilot of AI technologies in core voting systems at polling stations:

• Chatbot: On the CEC’s website, the Commission has integrated a chatbot with large language model (LLM) capabilities for responding to incoming voter queries about election logistics. The AI employs a hybrid model of closed- and open-form questions. Closed ‘decision trees’ prompt the user to select the category most closely aligned with their question, helping to triage users and providing information that is relevant to the specific conditions of each case. Once users have selected the appropriate categories, they are prompted to ask a follow-up question in free-form text.
• Image analysis for vote tabulation: Following the standard procedures for introducing new technologies, the CEC ran an initial pilot test of a new ballot-scanning software to support vote tabulation and results transmission during the 2025 partial mayoral elections. One of the five municipalities conducting elections was selected for a trial using this automated vote-counting system, which deployed image-analysis techniques to scan ballots in parallel with the manual counting process.
   

Areas where AI tools are currently under consideration

The CEC is evaluating various approaches to enhancing operations through the integration of AI tools. Although efforts are still in their early stages, initial discussions are focused on developing comprehensive, institution-wide solutions rather than limiting efforts to individual productivity or workflow management tools. However, in parallel with these large-scale efforts, the CEC continues to assess how established tools can be better leveraged to expedite and improve operations.

• AI-powered election day helpdesks for polling station practitioners: The CEC is training its chatbot (currently used only for public inquiries) to support internal coordination and communication. Specifically, the CEC is looking into how the LLM could be utilized to address rudimentary questions from polling station workers on election day and to alleviate the capacity strain placed on centralized human-led helplines.

• Transcription of public Commissioner hearings: As per ongoing discussions, the CEC is considering using AI transcription tools to help with minute-taking during public hearings with the SEC, the RC, and the Claims Commission. Currently, all notes are taken manually, so the introduction of AI would constitute full automation of the process.

   

   

Piloting and human oversight: Lessons from the vote tabulation pilot

The initial piloting of AI-powered vote tabulation systems in one 2025 mayoral election delivered a lesson on the importance of human oversight, particularly when AI is employed in core operations of electoral administration. Polling station staff used the image analysis system, which involved holding up a ballot that the system scanned and recorded for tabulation. After the system generated initially promising results by logging ballots with consistent accuracy, oversight personnel soon encountered two issues with serious implications for the system’s reliability.

First, polling station staff were often tired by the time vote tabulation started, meaning that they hurried when scanning individual ballots. Owing to quirks in the AI programming, the speed at which vote counters were holding up the ballots led the system to record each vote twice. This issue was only resolved when personnel were stationed to monitor carefully the vote counters to ensure that they held up each ballot long enough for the system to register it only once.

Second, owing to the rigid parameters of the image analysis system’s programming, it struggled to record ballots with minor discrepancies that would otherwise have been interpreted as a valid vote by human practitioners, effectively disenfranchising voters. This issue arose not only because of minor differences in how the ballots were filled in—such as slightly oversized check marks—but also because of the conditions in the polling station. On election night, the physical handling of ballots eroded marked lines on the tables that were used by the image-scanning system, reducing the system’s ability to register votes that were otherwise valid.

Reflecting on these experiences, the CEC stressed that AI implementation should not be based solely on theoretical risk prediction and preemptive contingency planning. Rather, careful piloting under human supervision is essential to pinpoint unforeseen outcomes that only become apparent during real-world testing, allowing electoral management bodies (EMBs) to address risks before they jeopardize the integrity of electoral administration.
 

Political acceptability and AI skepticism

One of the major challenges facing the CEC in the deployment of AI tools is the persistent controversy surrounding the digital automation of election administration. While the regulatory frameworks in place to test AI technologies serve a vital function in limiting the potential harm of any tool, even successful pilots can be received with a level of skepticism. Issues of public perception and AI acceptability arose during the pilot test of the vote tabulation system during the 2025 mayoral election. Representatives of political parties encouraged the CEC to delay releasing the AI-assisted results to avoid generating distrust in the technology. While polling station workers encountered technical issues during the initial tabulation process—as described above—the final system output displayed a near 100 percent congruence with the manual count results. This illustrates that, regardless of the technical successes of such systems, the automation of core voting processes requires high benchmarks for transparency and audit to reassure the public of their reliability.
The Commission’s response to this challenge follows a structured, mandatory process to test public acceptability. Under the Albanian Electoral Code, once the RC has taken a decision to pilot a new technology, the CEC is legally obligated to communicate the project to the public and conduct a public acceptability hearing (https://cdn.osce.org/sites/default/files/f/documents/2/2/159501.pdfAlbania 2008). This open hearing process is inclusive by design, inviting political parties, civil society, the media, and private citizens to engage with the technology under consideration. As a procedural mechanism, this dialogue-driven, participatory step ensures that technical innovation at the CEC is democratically legitimate and that system use is transparent.

AI policy development and the CEC strategic plan for 2027

So far, the CEC’s approach to AI governance has taken the shape of discrete policy instruments, tailored to each AI use case, rather than a unified policy framework. These regulations establish operational standards, define roles and responsibilities, and set out terms for managing voter data in accordance with the technical requirements of each application.

However, with the drafting of a new strategic plan for 2027 and beyond, the CEC is developing an overarching framework for the responsible integration of AI at an institutional level, with broader goals and evaluation metrics. The new strategy will include provisions on digital literacy and staff training, recognizing that building internal capacity and technological understanding is essential to mitigating adverse consequences such as those encountered in the vote tabulation pilot. These institution-wide regulatory efforts at the CEC align with the Albanian government’s broader commitment to develop AI projects and policy. The country’s intersectoral Digital Agenda 2022–2026 prioritizes the expansion of AI and advanced digital technologies across public services, highlighting their potential for raising the quality and efficiency of state institutions (National Agency for Information Society 2022). This is further operationalized through the Council of Ministers’ Decision No. 479, which establishes a national methodology and technical standards for the use of AI in the public sector, providing agencies such as the CEC with a reference framework for responsible AI deployment that draws directly from the European Union Artificial Intelligence Act 2024 (EU AI Act) (Parliament of Albania 2023; Bode and Xhaferaj 2026).