Argentina’s Security Ministry has launched an artificial intelligence unit that it says will “predict future crimes and help prevent them.” 

The announcement has been described as a real-life version of Minority Report, the 2002 Steven Spielberg film about a futuristic policing unit that detains criminals before a crime has been committed. It has also raised concerns about state surveillance, the social biases AI often reflects, and the societal tendency to significantly overestimate the powers of products caught in the latest tech gold rush.

The AI unit’s launch comes as the government says it wants to make Argentina an “AI hub.” Meanwhile, Congress is only starting to discuss the implications of the technology — and how the country might regulate it.

The new unit will use AI to patrol social media, the internet, the dark web, apps, and security camera footage, according to the resolution published in the official gazette on July 26. Signed by Security Minister Patricia Bullrich, the resolution states that the unit will use machine learning algorithms to analyze historical crime data, so it can “predict future crimes and help prevent them.” It also plans to fly surveillance drones and use AI to analyze the footage.

The unit will be led by the director of the Security Ministry’s cybercrime area, Federico Pierri.

Ministry spokespersons contacted by the Herald did not confirm what software the unit would use, or give specifics on where the data to be analyzed would come from.

‘Selling snake oil’

Beatriz Busaniche, director of digital rights defense foundation Vía Libre, said the project appeared to reflect a general trend of “using AI on everything without having a clear understanding of what AI means.”

“There are a lot of different technologies that are all packaged under the label of AI,” she said. “They are selling snake oil.”

She warned that “cyber patrolling” constitutes “illegal surveillance,” since it is illegal to conduct intelligence activities on citizens without a warrant in Argentina.

AI has a patchy record when it comes to identifying possible troublemakers — and the errors often reflect the discriminatory biases of the society that created it. A 2016 investigation found that COMPAS, a risk assessment tool used in the United States to decide whether prisoners should be given parole, was biased against African Americans.

It can also be used as a weapon: this year, it was reported that the Israeli army identified Palestinians in Gaza as bombing targets using Lavender, an AI-targeting system.

There are also disturbing questions about how government political bias could interact with an AI security system, according to AI engineer Matías Gringberg of HUMAI, an NGO that promotes the use of AI for social good. For example, President Javier Milei’s government accused people who protested in June against the Ley Bases reform package of an attempted coup d’etat.

“I would consider it unacceptable unless there is an ethics committee that is trans-organizational overseeing every action and development,” Gringberg said.

He is also concerned about data protection, since Argentine government institutions have been hacked in recent years. When the RENAPER national persons registry was hit in 2021, the national identity card numbers of everyone in Argentina were leaked online.

Facial recognition

“This would be more sensitive data, as it is personal and would allow people to be identified through facial recognition on a much larger scale,” he said.

Buenos Aires already has a facial recognition system, which was installed in 2019 during the Mauricio Macri presidency — and it has been abused. In 2022, a judge ordered for the system to be suspended after was found to have been used for over 9 million biometric data searches, far more than the 40,000 criminals that officials claimed the system was targeting. The targets of these searches included Alberto Fernández and Cristina Fernández de Kirchner, who were president and vice president at the time.

Technology exists to allow the government to set up a vast web of monitoring and information gathering, but how that would translate into robust and accurate crime prevention is less clear, according to Grinberg. “It would be a great sacrifice of human rights to, if they’re very lucky, perhaps identify or prevent a crime,” he said.

Bullrich’s announcement came a week before a Lower House commission started to discuss several bills to regulate AI. On Sunday, Demian Reidel, chief advisor of President Milei, reiterated his desire to turn Argentina into an “AI hub.” He was dismissive of proposals to regulate the technology.

“We have to be very careful not to start making stupid regulations that are going to hurt [the prospect of investment],” Reidel said on the Club de Inversores streaming channel. 

“No stupid regulation is going to pass.”