From algorithm to vote: how to protect democracy in the age of AI?
Diana Dajer, Democracy Manager at Fundación Corona.
On July 28, the Ministry of Science, Technology and Innovation (MinCiencias) submitted a new bill to regulate artificial intelligence (AI) in Colombia. The bill represents progress, albeit with challenges , in the protection of fundamental rights. With this proposal, there are now more than ten legislative initiatives on AI currently being processed in Congress, but none have focused on defending democracy and electoral processes .
And what about democracy? Amid the regulatory boom surrounding AI, no one seems to be talking about it . While Congress is piling up bills to regulate this technology, AI is already transforming how politics is conducted: how we stay informed, how we participate, and how we make decisions. Ignoring this impact in regulatory design could leave essential pillars of the democratic system unprotected, such as access to accurate information, equality in public debate, and transparency in elections.
Democracy cannot be a collateral damage of technological advancement. In this article, I will discuss the three main risks that AI poses to Colombian democracy and offer concrete recommendations that could be incorporated into the regulatory framework, public policies, and complementary actions to protect it. This will be based on various sources on AI in democratic contexts, such as the database on AI regulation in Colombia and Latin America from the School of Government at the University of the Andes, and documents from the Alan Turing Institute , the AI + Society initiative at the University of Ottawa , Rest of World , and the German Marshall Fund.
1. Disinformation, polarization, and hyper-personalized manipulation
Artificial intelligence allows for the creation and dissemination of fake political content on a massive and personalized scale. For example, in Colombia, during the 2023 elections, fake AI-generated audio recordings circulated , impersonating candidates Carlos Fernando Galán and Alejandro Eder. These recordings spread rapidly through social media and chat groups, damaging the credibility of their campaigns and confusing voters.
AI enables political micro-segmentation: messages tailored to each voter based on their interests and online behavior. While this technique is not inherently negative, it can violate fundamental rights if it relies on the mass use of personal data without consent, incites hatred , or is used to make disinformation more effective.
Cases such as Cambridge Analytica in Brexit and the 2016 US presidential elections or the use of deepfakes in the 2024 Indian elections show how AI can manipulate individual preferences on a large scale, eroding trust in the electoral system.
In Colombia, Law 2502 of 2025 classified identity theft using AI as an aggravating criminal offense and mandated the creation of an intersectoral public policy to address its risks. However, much remains to be done, and some measures must go beyond mere regulations, such as:
Require that all AI-generated political content be clearly labeled.
Promote transparency in the use of electoral segmentation algorithms.
Strengthen the capacities of the National Registry and the National Electoral Council to monitor and sanction the misuse of AI.
2. Attacks on electoral infrastructure using AI
Artificial intelligence can also be used to attack the systems that underpin our elections : voter databases, official websites, registration platforms, and results transmission systems. For example, according to the Brennan Center for Justice , AI can simulate denial-of-service attacks, manipulate user interfaces, and generate fake content that mimics official communications. These attacks can sow confusion and distrust in the electoral process; likewise, in highly polarized contexts, they can fuel narratives of fraud.
There are already concrete examples. In the 2024 US elections, AI-generated fake messages were reported telling voters in key states that they had already voted or that their registration had been canceled. These messages aimed to confuse and discourage voter participation. In the UK, in 2023, a cyberattack on the Electoral Commission compromised the data of millions of voters, which could facilitate targeted disinformation campaigns.
In Colombia, these risks have not yet been comprehensively addressed in the bills to regulate AI. Some specific recommendations to protect electoral infrastructure are:
Educational campaigns to help citizens learn to identify official communications, protect their data, and recognize attempts at digital manipulation.
Monitoring, early warning and rapid response systems to possible automated attacks, in collaboration with national and international cybersecurity centers.
A specific strategic line on digital electoral security within the public policy mandated by Law 2502 of 2025, with resources, technical capabilities and specialized training.
3. Fraud against rules on the financing of electoral campaigns
Since 2020, the National Electoral Council in Colombia has recognized that the use of social media for electoral purposes constitutes political propaganda. Digital platforms have opened new ways to advertise political content, but they have also created mechanisms to conceal who pays for these messages and to publish them outside the permitted timeframe for electoral campaigning.
For example, in Mexico , the National Electoral Institute ordered in 2021 the removal of a campaign by the Green Ecologist Party of Mexico that was being disseminated through influencers on social networks, for carrying out propaganda in a prohibited period.
In Romania , the Constitutional Court annulled the first round of the 2024 presidential elections after a Russian-funded hybrid attack campaign was revealed, which favored pro-Russian candidate Calin Georgescu through cyberattacks and manipulation on social media such as TikTok.
AI allows for the creation of fake identities, the simulation of donors, and the movement of money through bot networks and automated accounts in a way that is very difficult to detect. To address these risks, the State should implement specific measures aimed at strengthening the transparency and traceability of digital political financing, for example:
Require digital platforms to report who pays for and how political advertising is targeted .
That the National Electoral Council has better tools to track digital money.
Conclusion: Protecting democracy in the face of disruptive technology
Artificial intelligence is rapidly transforming democratic processes. Law 2502 of 2025 and the new bill presented by the Ministry of Science are important steps forward, but the protection of democracy must be an explicit focus of any regulatory framework on AI. Likewise, as seen in the various recommendations proposed, this protection must, in many cases, extend beyond the regulatory sphere. Voting is still a human act. Protecting it in all its dimensions is an urgent task.
See the article on La Silla Vacía here: https://www.lasillavacia.com/red-de-expertos/red-de-democracia-y-tecnologia/del-algoritmo-al-voto-como-proteger-la-democracia-en-la-era-de-la-ia/