The use of Artificial Intelligence (AI) can improve the efficiency of elections, from pre-election preparation to post-poll procedures. At the same time, however, AI use brings technical and political vulnerabilities that pose real threats to democratic systems. Preserving electoral integrity during and after AI integration requires a multifaceted approach, combining institutional, technical, and legal safeguards against manipulation by state and non-State actors throughout a model's operational lifecycle. A holistic policymaking approach helps identify technical threats and countermeasures, enforced by designated AI committees with diverse membership to ensure ongoing human accountability. These measures should be underpinned by new legislation mandating algorithmic transparency, oversight, and shared responsibility among stakeholders. This brief outlines the preventive steps needed to integrate AI securely into electoral systems and safeguard their integrity.
The administration of elections is an area where AI tools have potential for widespread use. Indeed, in different parts of the world, democratic exercises are becoming progressively embedded within a broader socio-technical ecosystem shaped by data extraction, user profiling, and predictive analytics. As the applications of AI in electoral contexts continue to expand in parallel with advances in the technology itself, it is essential to examine this integration and to curb its potential as a threat vector to the very same democratic processes that it purports to aid. Even as the efficiencies realised through the deployment of AI in electoral procedures—including voter registration, results tabulation, and civic engagement initiatives—are well documented, the extended use cases and associated threat landscape must be mapped.
This is particularly pertinent given that AI technology, especially when applied directly to democratic exercises, is still in its initial stages of deployment and has been limited mainly to low-risk applications. Consequently, the majority of threats currently identified as risks to electoral processes remain largely hypothetical, with limited precedent to substantiate existing concerns. However, given the sensitivity of electoral politics and the foundational role of elections in legitimate governance, it is necessary to address all plausible risks in advance.
{{/usCountry}}This is particularly pertinent given that AI technology, especially when applied directly to democratic exercises, is still in its initial stages of deployment and has been limited mainly to low-risk applications. Consequently, the majority of threats currently identified as risks to electoral processes remain largely hypothetical, with limited precedent to substantiate existing concerns. However, given the sensitivity of electoral politics and the foundational role of elections in legitimate governance, it is necessary to address all plausible risks in advance.
{{/usCountry}}This paper can be accessed here.
(The views expressed are personal)
This paper is authored by Pranoy Jainendran, ORF.