Democracy vs. Deepfakes: Latin America’s Electoral Stress Test
Argentina’s latest AI scandal shows that fake politics is becoming a real threat.
The release of a deepfake video of the governor of Argentina’s Córdoba province in the run-up to the country’s legislative elections has fueled renewed discussion of how to appropriately regulate AI during elections.
Córdoba Governor Martín Llaryora was depicted in the video giving a speech in which he says that good governance is for “dumbasses” because campaigns should focus on finding people to blame rather than finding solutions, and that legislators should blame the government of President Javier Milei for their own shortcomings.
Córdoba legislator Rodrigo de Loredo said he created the video “based on real events, written and spoken by AI” and called it an accurate description of the Peronist political movement.
Journalists discussing the issue on Infobae noted that the disclaimer does not eliminate confusion about its authenticity because the video was then distributed on social media accounts that omitted the explanation that it was fake.
In response, one legislator proposed changing the country’s elections code to block candidates from participating in elections if they seek to manipulate the electorate through digital deception.
De Loredo has not taken the video down, but offered to do so as long as Governor Llaryora lowers taxes.
Free speech or misinformation?
The incident comes months after Milei allies released an AI-generated video showing Argentina’s former President Mauricio Macri falsely announcing the withdrawal of a candidate for the Buenos Aires legislature (the candidate, Silvia Lospennato, lost the election by a wide enough margin as part of a landslide for Milei allies.)
Milei called Macri a “crybaby” and defended the release of the video as free speech and part of the new forms of communication by young people.
The incident fueled discussion in Peru of how the AI deepfakes could influence the 2026 general elections.
Chile, which has presidential and parliamentary elections coming up this year, is also grappling with how to manage the impact of artificial intelligence.
Is Latin America ready?
The short answer appears to be no - most countries do not have established AI regulations or clear measures to address the potential electoral impact of deepfakes.
Most draft legislation in the region focuses on labeling AI content and requiring disclosure when campaigns use AI for text and voice generation, but few criminal penalties for real-time enforcement mechanisms to pull viral content from circulation.
Chile, which will hold general elections in October, is currently in the drafting stage with a focus on ethical AI development with only a preliminary discussion of electoral integrity.
Brazil has the most robust electoral protections, with an express prohibition on the use of AI deepfakes in elections and Brazil’s Congress this week discusses legislation to regulating all AI content - not just deepfakes - and to create mechanisms for requiring that content be removed.
But the existing prohibition did not prevent deepfakes from appearing in the 2024 municipal elections, according to a report by the Digital Forensic Research Lab (DFRLab) at the Atlantic Council, which included pornography deepfakes targeting female candidates and content falsely accusing candidates of crimes.
Upcoming elections in Bolivia and Uruguay will provide good indicators of how the region manages AI’s impact on voting.
“Get Up to Speed, Idiots” - When AI Outsmarts Accountability
Mexican Legislator Olga Chávez Rojas of the ruling Morena party acknowledged using artificial intelligence to summarize legislative proposals to avoid reading hundreds of pages herself. She called it a legitimate use of new technology and told legislators from the PRI party “Get up to speed, idiots.” (El Financiero)
Her comments triggered criticism that she was delegating legislative deliberation to AI to rush through citizen security reforms without sufficient consideration for their civil liberties implications, including the use of facial recognition software. (Infobae)
In all fairness, it’s not uncommon for legislators to have staffers read long and tedious documents and receive summaries. This is not the same as tasking this out to an LLM, but also not the same as reading it yourself. And ramming legislation through congress carries its own risks whether or not AI is involved.
Macro Prompt
Telltale Chatbot Verbiage
Biomedical abstracts published in the last two years show an increasing use of a specific group of words such as “delves,” “crucial,” “potential,” “significant” and “important” - verbiage favored by chat bots, according to research published in the journal reports the New York Times.
“At least 13.5 percent of all biomedical abstracts appeared to have been written with the help of chat bots. And as many as 40 percent of abstracts by authors from some countries writing in a few less selective journals were A.I.-generated,” the story reads.
How many of these researchers are not native English speakers? Annoying AI parlance seems like a small price to pay to help researchers around the world spend more time on results and less time fussing over their English. The line here seems to be whether AI is misrepresenting the results, or doing the science itself.
Yes, You’re Talking in Robot Voice
Human beings are starting to talk - out loud - like chatbots, according to a study at the Max Planck Institute for Human Development.
Researchers who studied 750,000 hours of human discourse since the release of ChatGPT observed “a measurable and abrupt increase in the use of words preferentially generated by ChatGPT—such as delve, comprehend, boast, swift, and meticulous.”
“This marks the beginning of a closed cultural feedback loop in which cultural traits circulate bidirectionally between humans and machines … rais(ing) concerns over the erosion of linguistic and cultural diversity, and the risks of scalable manipulation.”
I was unfortunately unable to come up with an annoying ChatGPT catch-phrase to evidence this shift. I’ve definitely noticed some circuitous answers and tendency for repetition. I’m not, however, going to give up em-dashes — no matter how much the bots like them.
Cleaning Up AI blather
The business of writing is back: companies are looking for copy editors to fix poorly written AI-generated content, reports the BBC in a story aptly entitled 'I'm being paid to fix issues caused by AI'
China’s AI Push
Chinese artificial-intelligence is challenging American dominance, with startups such as DeepSeek and e-commerce giant Alibaba as alternatives to American companies such as ChatGPT, reports the Wall Street Journal.