BA province regulates state AI use after chatbot named as an official

Buenos Aires province has become the first district in Argentina to regulate the use of artificial intelligence in government, launching its rules just two weeks after a city appointed a chatbot as a public official. Earlier this month, Zárate mayor Marcelo Matzkin appointed a chatbot named ZARA as “director general of non-human citizen services.” Official paperwork calls ZARA “a non-human civil servant” working at the undersecretariat of innovation and digital accessibility.  Intended to cut waiting times and expand access to services, the bot’s decisions, communications, and responses are to be considered legally and administratively binding, according to the decree naming ZARA to the role. The unusual appointment — and the provincial government’s decision to regulate — put the spotlight on Argentina’s national policy on the issue. President Javier Milei’s administration has so far argued that regulating AI would scare off investors. Matzkin, from the conservative PRO party, told newspaper La Nación that the bot was intended to automate procedures such as the authorization of applications to operate retail businesses in areas where they were already permitted. “We want AI to analyze what you submit and grant you authorization with administrative validity,” Matzkin said. Last Thursday, two weeks after ZARA’s appointment, the Buenos Aires provincial government issued regulations on the use of AI in government.  Sandra Dagostino, Buenos Aires province’s digital government undersecretary, told the Herald that the regulation had been in the works for months and was not related to the creation of ZARA. However, she added that she disagreed with the appointment of an AI tool as a municipal official. The new system is aligned with the European Union’s AI Act. It classifies technologies based on risk, ranging from zero risk to unacceptable risk. Examples of zero-risk uses are systems such as spam filters, which pose virtually no risk to security or fundamental rights.  Unacceptable risks include AI tools that pose a threat to health and safety, fundamental rights, and the democratic order, among other perils. Examples include systems capable of exploiting vulnerabilities of a particular group, profiling people according to their behavior, or inducing dangerous behavior change through subliminal techniques. Officials who want to use AI systems in government will now have to submit them for assessment before implementing them.  “Decisions can’t be totally automated,” Dagostino said. “Every high-risk action needs human control.”  The Buenos Aires provincial government is led by left-leaning Peronist Governor Axel Kicillof. Dagostino said the government is not anti-AI and uses the technology for tasks such as classifying interactions on government hotlines. However, they also consider issues such as data protection and energy consumption when choosing technology. Sometimes, more traditional tools are a closer fit. “Not everything requires AI,” she said.