Prometea — Buenos Aires' AI for judicial ruling prediction and prosecution automation

Prometea was developed from 2016 by the Ministerio Público Fiscal of the Autonomous City of Buenos Aires in partnership with the Universidad de Buenos Aires Law School. It uses natural language processing and machine learning to analyse case precedents, predict judicial outcomes with stated 96% accuracy, and automatically draft preliminary ruling documents.

To speed up case processing and drafting of preliminary rulings while keeping prosecutors as the final decision-makers, with AI output explicitly advisory.

In its first major deployment (2017-2018), Prometea processed 1,000 labour-rights rulings in 5 days compared with the prior manual average of 83 days; in housing-rights cases processing time fell from 174 days to 45 days. Prometea is open-source (MIT licence). The Colombian Constitutional Court adopted the system for case-selection workflows, and the Inter-American Development Bank has provided technical assistance for further regional deployment.

These figures are documented in the Inter-American Law Review (University of Miami) and cited in UNESCO's 2023 AI and Rule of Law Toolkit. The Colombian Constitutional Court reduced the time to shortlist a case from 96 days to 2 minutes, demonstrating cross-border legal-system transferability.

The full training dataset and model weights are not publicly released, limiting external reproducibility of the 96% accuracy figure, and no formal independent algorithmic audit has been published, though UNESCO endorsement and IDB involvement imply external validation at a high level.