Why we should all be worried about AI in elections

In the first week of March, 193 United Nations member states convened in Geneva for the inaugural Global Dialogue on AI Governance. However, the discussions largely overlooked a critical issue: the impact of artificial intelligence on elections. This omission could undermine advancements across all other aspects of the AI debate, including safety, trustworthiness, capacity building, societal implications, and human rights.

Elections are the primary mechanism through which other democratic rights are enforced, yet efforts to consider AI's role in elections often focus on disinformation and deepfakes. Yet, preliminary reports from the U.N.'s Independent International Scientific Panel on AI highlight these concerns without addressing the more significant risks AI poses to the systems that administer elections.

A prime example is India's experience with biometric voter-verification systems. In 2015, the Election Commission of India integrated voters' photo ID records with Aadhaar, the biometric national identity database. This machine learning algorithm flagged and deleted up to 93% of duplicate and dead voters without providing any notice or re-enrollment options. Subsequently, the system was revived nationally in 2021, but the flawed algorithm continued to pose risks to India's electoral integrity.

The implications extend beyond individual countries. In many African, Asian, and Latin American nations, election technologies are often designed, supplied, and hosted by private vendors operating outside national electoral oversight. These AI-enabled systems can compromise sensitive voter data, leave proprietary systems unauditable, and introduce hidden algorithmic errors that could disproportionately exclude marginalized communities.

AI's opacity and complexity amplify these risks. Without clear governance standards, election authorities may find themselves accountable for decisions embedded in technologies they cannot fully inspect or regulate. This underscores the need to foreground elections in global AI governance discussions, ensuring that AI does not undermine democracy's institutional foundation.

The African Union's recent mandate to shape and control its AI ecosystem highlights a promising approach. By anchoring AI governance in regional realities and mandating conditions like data localization and technology transfer, the AU Advisory Group on AI in peace, security, and governance sets a precedent worth replicating globally. Addressing AI's governance in elections is crucial to maintaining the public's right to participate in the democratic process and holding leaders accountable.