The tech narrative in the last two years has been dominated by AI and the excitement and disruption generated by it. However, the narrative started to sour a little in the latter half of 2023, as people started getting uncomfortable about the ethical issues thrown up like copyright, bias, privacy, and above all, deepfakes. Now, with elections looming across most of the democratic world, 2024 promises to be the year when AI will experience its first major ethical test – on whether it can help democracy or help subvert it. A set of circumstances have created a perfect storm this year. Major democracies like India, the US, UK, Indonesia and others go for pivotal elections this year. While deepfakes existed before GenAI, products like Sora and Stable Diffusion have democratized the production of deepfakes, making them easier, faster and cheaper to make at scale. We are also at peak social media, with TikTok, WhatsApp and the like making their global distribution a piece of cake. Bangladesh and Slovakia were the first countries that went to polls this year, and expectedly deepfakes came to the party. A Bangladeshi Opposition leader was shown to be ambivalent about his support for Palestinians, a disastrous position to take in the country. In Slovakian elections, a major contender reportedly talked about rigging the elections, and more alarmingly, raising the price of beer, reportedly contributing to his defeat. A fake voice of President Biden urged people not to vote in the US Primaries. With the memories of the 2016 Cambridge Analytica debacle still fresh, these early signs have set off alarm bells as the big elections loom later this year.

This is where I take a contrarian stance and urge you to look at Pakistan. The country went to the polls recently, with a former Prime Minister in jail, his party symbols taken away, and party candidates threatened and imprisoned. While eventually other parties were declared to have won, most reports claim that Imran Khan’s party won convincingly despite heavy rigging and manipulation. Khan turned the narrative of AI subverting democracy on its head, by leveraging GenAI to canvas across the country, while behind prison walls. GenAI was used to create footage of his urging voters to come out and vote for his party, and this was widely shared on YouTube and other social channels. The people heeded his call and came out in record numbers, handing over astonishing victories to his candidates. Without necessarily taking sides here, Pakistan showed a way how AI could be used in a manner different from the prevalent narrative of a democracy-destroyer.

I am not denying the destructive power of deepfakes, and I fear their use in Indian and other elections to inflame discourse and shape narratives. However, there is a lot of good that AI can also do to improve elections, a core pillar of democracy. The Pakistan example is one creative way, but there are others. AI can be used to enhance transparency, inclusivity, and efficiency in elections. Advanced data analysis capabilities of AI can monitor and analyse election-related data in real-time, identifying any irregularities suggesting fraudulent activity. AI algorithms can detect patterns of irregularities in voter registrations or ballot submissions, identifying potential concern areas. AI can also improve the security of electronic voting systems, a worry of many opposition parties. Additionally, threat detection algorithms can also help identify and counter potential cyber threats. Generative AI can help uplevel voter education and awareness by generating hyper-personalized content on candidates and their manifestoes, in their local dialect pertaining to local constituency issues. This personalized approach could enhance political awareness and informed voting, especially among marginalized or disengaged groups. GenAI can help do this at scale with much lower costs and higher efficiency, thus empowering even the less moneyed candidates. AI-powered systems can enhance accessibility for voters with disabilities to increase inclusivity among the electorate. For instance, AI-powered voice

recognition systems can assist visually impaired voters in casting their votes independently. AI can parse the firehose of information available in social media to gauge public opinion across multiple demographic groups, ensuring that all sections of society are represented in the political conversation. Even something as mundane as the logistics of electioneering can be optimised and costs saved, something very important for large countries like India. AI can help make voter registration and verification more efficient, and also help improve voter experience, eliminating long queues providing all information required to vote beforehand.

I have written before about how AI is a dual used technology, with huge benefits accompanied by immense destructive power. As we try and contain the adverse effects of AI on elections through deepfakes, we should also look at how we can use it to improve our fragile democracies too. While not in a perfect way, Imran’s party has succeeded in pointing this out for humanity.