Artificial intelligence has stormed the world of democracy in two significant ways. First, it helps governments win elections. Second, it helps them govern after victory. However, neither process is monitored, and the rules meant to keep them in check are far behind. For example, political candidates and parties have used political ads and videos made by AI because they are inexpensive to produce. A proposal on deepfakes was met with concerns that it would erode people’s trust in the electoral system itself.
The bigger issue is not the fake clips, but the loss of trust. People around the world are concerned that politicians can now claim real evidence as fake simply because fakes are the new normal. People think that, by knowing that fakes are possible, they can label anything that causes them inconvenience as fake, even if it is true or authentic. This has now come to the attention of lawmakers, who are trying to address the issue. However, their responses have been mixed. Many states have adopted AI regulations that require ads to be labelled as AI-generated. However, it is uncertain whether these new laws can prevent the problem from worsening or eliminate it entirely. California attempted to take a more significant step by banning deepfakes, but a federal court overturned the ban on the grounds that it violated fundamental rights. For now, what is needed is a consistent, strong rulebook.
The problem is just as bad in other countries, regardless of religion or type of governance. During India’s 2024 elections, for example, YouTube and WhatsApp were flooded with fake videos of Prime Minister Narendra Modi campaigning for the opposition party. However, researchers say that it has not been proven whether these videos impacted the election results. They also confirmed that a constant surge in fabricated content would have an adverse effect on how people perceive political standings, even if such fakes did not influence votes.
Pakistan has demonstrated how quickly situations can escalate. During the 2024 general elections, an AI-generated clone of Imran Khan was used by his party to deliver victory speeches while he was in prison, causing widespread debate among the 128 million voters. Similar AI-driven rumours were reported in Bangladesh and Indonesia during their 2024 elections. One such rumour involved an AI-generated audio clip that attempted to influence voters before the polls opened. This problem transcends borders and income levels.
An even greater problem is the extent to which AI is being used to assist the government in making everyday decisions. If used carefully and intelligently, AI could help the government make faster and fairer decisions about who receives welfare benefits or tax exemptions and who is investigated. However, it can create a system with no recourse. The Netherlands’ government developed an algorithm called SyRI to detect welfare fraud, but it became an inexplicable black box that violated people’s right to privacy, and it was eventually rejected by a Dutch court.
The OECD admits that governments emphasise the importance of building trustworthy AI, but have not been successful in creating the necessary tools. Some authoritarian countries have clearly made their choice. Freedom House’s latest report reveals that internet freedom has declined for 15 consecutive years, partly due to AI-powered surveillance and censorship. At least 22 countries now legally require platforms to remove speech against the government. Worryingly, even countries rated as ‘free’, including the US, saw a decline in their score this year.
As Bruce Schneier, a Harvard-based security researcher and fellow at the Berkman Klein Center, put it, the present moment is the most AI-driven that anyone in the field can remember — not because it was deliberate, but because technology arrived faster than rules. The real question of concern is not whether democracy will be shaped by algorithms, but whether democracies will be able to establish the rules first or learn the hard way.