The article argues that the biggest threat AI poses to elections may not be deepfakes or viral political misinformation, but the quiet integration of AI into the systems that actually administer elections. The authors, Lisa Poggiali and Samson Itodo, say voter-registration databases, biometric verification systems, election-results software, cloud infrastructure, and other election technologies increasingly incorporate AI. Because these systems operate behind the scenes, errors or biased decisions may affect voters without attracting the public attention that a fake video of a politician would.
India is presented as an important example of these risks. The article discusses efforts to link voter records with Aadhaar and use machine learning to identify duplicate or deceased voters. According to the authors, millions of voters were removed from electoral rolls in two states before India's Supreme Court halted the program, while subsequent disclosures reportedly showed very high error rates in the matching system. The example illustrates the potential danger of allowing an algorithm to influence who remains eligible to vote, particularly when affected citizens may not know why they were removed or have an effective way to challenge the decision.
The authors also warn about the growing role of private technology vendors and AI chatbots in elections. Election authorities in many countries may rely on systems that they did not build and cannot fully inspect, while voters increasingly turn to tools such as ChatGPT and Gemini for political or election information. AI-generated answers can contain factual errors or reflect biases, and these problems may be particularly serious in countries with linguistic diversity or limited reliable digital information. The concern is therefore not only misinformation deliberately created by political actors, but also ordinary AI systems unintentionally giving voters incorrect information.
The article ultimately calls for elections to become a central part of global AI governance. The authors argue that governments should be able to determine who controls electoral data, independently audit AI systems used in elections, and establish clear responsibility when an algorithm affects someone's voting rights. They point to an African Union approach emphasizing data localization, technology transfer, and disclosure of source code as potential safeguards. Their broader message is that protecting democracy requires protecting the institutions through which citizens exercise political power—and that means ensuring AI systems used in elections remain auditable, accountable, transparent, and under meaningful public control.
