"For twelve hours every crime within the limits of our country will be absolutely legal", says the alleged voice of Javier Milei, presidential candidate in Argentina, in a video that has circulated during the last few days in social networks. According to the video, which at times shows the candidate explaining the proposal, under his government the plot of the movie The Purge ( 2013), in which a dystopian regime authorizes a day for citizens to commit crimes with total impunity, would be put into practice.
The piece is a deepfake, an artificial intelligence (AI) technique that allows manipulating people's voices and faces to create fake content. Despite the fact that the far-fetched discourse and the use of images from the movie allow inferring that it is not a real proposal, the video managed to circumvent Meta and YouTube's review systems and was even promoted as political advertising in these social networks.
In theory, advertising on platforms is subject to higher levels of supervision than organic content. However, according to information from the Meta Ads Library, advertising about Milei's alleged purge was seen between 250,000 and 300,000 times on Facebook and Instagram.
Meta's policies prohibit AI-manipulated videos that would make an average person believe that the person in the video "utters words they did not utter". These criteria may be too narrow to combat potential negative uses of artificial intelligence.
On the one hand, the rules are limited to videos, thus excluding images created through generative models such as Dall-E or Midjourney. On the other hand, it leaves out other types of manipulation that do not necessarily involve these technologies, as recently pointed out by journalist Casey Newton and as is likely to be reviewed by Meta's Content Advisory Board in the wake of the case of a video altered to label Joe Biden a "pedophile."
The advertisement of Milei's purge was deactivated by Meta for not having a disclaimer, a requirement established by the company for advertising related to social, political and electoral matters to indicate the entity or person in charge of publishing the advertisement. Behind the advertisement is Alerta Argentina, a page with few followers created in May of this year and which simply presents itself as a "business service".
In the case of YouTube, the same video appeared as advertising promoted by companies based in Hong Kong, as reported by users on social networks. Although Google has an Advertising Transparency Center, this tool does not include content removed for violating advertising policies or community standards, so it is not possible to track the strategies and actors involved.
The rise of AI and the democratization of access to generative video, audio and text models present new avenues for the dissemination of more sophisticated and compelling disinformation narratives. In Latin America, for example, these technologies have been used to spread deepfakes on Facebook and TikTok in favor of Andrés Manuel López Obrador's government in Mexico and to promote Nicolás Maduro's propaganda on YouTube and Venezuelan state media.
According to the Freedom on the Net 2023 report, recently published by Freedom House, in the last year AI has accelerated the deterioration of Internet freedom. Among others, generative models have been used to silence activists and propagate false theories of electoral fraud.
In addition to the irruption of AI, the electoral season in Latin America - which also includes the local elections in Colombia and the new constitutional plebiscite in Chile - has been the scene of more artisanal types of manipulation.
In the case of Colombia, for example, a video was published in recent weeks in which Federico Gutiérrez, candidate for mayor of Medellín, had allegedly been clandestinely recorded by a webcam model. However, it was a fragment of an interview made by an influencer and was decontextualized through subtitles.
A similar case occurred with Sergio Massa, presidential candidate in Argentina and current Minister of Economy, who appeared in a video receiving a transparent bag with a white content. According to the subtitles of the publication, when looking at the camera Massa said "I was always a merchant", suggesting that what he was receiving was cocaine. According to the agency Chequeado, which verified the piece, what he actually said was "I was always a goalkeeper", and the contents of the bag were papers.