Argentina's top court bans AI from dictating sentences, valuing evidence

The Supreme Court of Justice of the province of Buenos Aires has taken a significant step to regulate the use of artificial intelligence within its courts, establishing a precedent in Argentina. The court has explicitly prohibited AI from dictating sentences or evaluating evidence, emphasizing that judicial functions are non-delegable and irreplaceable.

This move comes as AI tools are increasingly used by legal professionals for tasks ranging from client assistance to legal research. However, the court acknowledged the risks associated with generative AI, including potential "hallucinations" or inaccuracies, biased outputs, and the "automation bias" where users overly rely on algorithmic suggestions.

“

the judicial function is non-delegable and irreplaceable

To manage AI integration, a new "Commission for Governance and Use of Artificial Intelligence" has been formed, comprising representatives from the Planning, Information Technology, and Judicial Services Secretariats. This commission will act as the implementing authority, tasked with proposing protocols, evaluating tools, conducting pilot tests, monitoring outcomes, and submitting regular reports.

The regulation strictly forbids AI from being used for predictive recidivism risk assessments that influence decisions on liberty or pre-trial detention. It also bans the creation of "synthetic" or automated evidence and the generation of expert reports without critical validation by a qualified expert. The court categorizes AI applications by risk level, with administrative tasks like transcribing hearings deemed "low risk." High-risk applications, such as those assisting in legal interpretation or predicting litigation outcomes, will require semi-annual audits and an "Algorithmic Impact Assessment" before implementation.

“

automation bias