Why we should all be worried about AI in elections

A global summit convened by the United Nations last month in Geneva brought together representatives from all 193 member states to address the growing need for international guidelines on artificial intelligence. Despite the broad scope of discussions, covering topics such as AI's safety, trustworthiness, and societal implications, the issue of AI's influence on democratic processes, specifically elections, received minimal focus. Experts warn that this oversight could weaken efforts to regulate AI effectively in other areas, including ethical considerations and human rights protections. The dialogue, known as the inaugural Global Dialogue on AI Governance, aimed to establish a framework for managing AI's rapid evolution. However, the absence of detailed discussion around AI's role in elections has raised concerns among scholars and policymakers. While the preliminary report from the U.N.'s Independent International Scientific Panel on AI touched upon AI's potential to spread misinformation through deepfake technology, it largely overlooked the broader systemic risks posed by integrating AI into the very mechanisms that uphold democratic participation. In recent years, the integration of AI into electoral systems has become increasingly common. Voter registration databases, biometric identification tools, and election result management software now rely heavily on AI algorithms. These systems are often developed and maintained by private companies operating outside the direct oversight of national electoral commissions. This shift raises questions about transparency, accountability, and the potential for misuse or malfunction within critical democratic infrastructure. India provides a notable example of these concerns. In 2015, the Indian Election Commission initiated a project to link voter identification records with the Aadhaar biometric national identity database. The initiative aimed to identify and remove duplicate or deceased voters from the electoral roll. However, the process relied on a machine learning algorithm managed by the Unique Identification Authority of India, which operates under executive control rather than as an independent entity. The outcome was alarming: approximately 5.5 million voters were removed from the rolls in two states without proper notification or means of appeal. A subsequent review found that the algorithm had a failure rate as high as 93 percent. Although the Supreme Court of India intervened and halted the program, similar initiatives have since been reintroduced nationwide, raising fears about the reliability and fairness of AI-driven electoral administration. Beyond specific instances of algorithmic error, there is a broader concern regarding the decentralization of electoral power. In many regions across Africa, Asia, and Latin America, private firms design, supply, and host election-related technologies. These entities often operate beyond the jurisdiction of national regulatory bodies, creating vulnerabilities related to data security, system audits, and the detection of algorithmic biases or failures. The increasing reliance on AI-powered chatbots for voter education further complicates matters. Platforms such as OpenAI's ChatGPT and Google's Gemini provide information to voters ahead of elections. However, these systems can generate misleading content or amplify existing biases, especially in areas with limited internet access or linguistic diversity. Such issues may disproportionately affect marginalized communities, exacerbating existing inequalities in political engagement and representation. As AI continues to permeate electoral systems globally, the challenge lies not only in addressing individual technological failures but also in ensuring that the structures governing elections remain transparent, accountable, and resilient to manipulation. The lack of comprehensive international standards for AI governance in elections underscores the urgency of developing robust frameworks that safeguard democratic processes against both intentional interference and unintended consequences of technological advancement.