The OECD.AI Policy Navigator

Our policy navigator is a living repository from more than 80 jurisdictions and organisations. Use the filters to browse initiatives and find what you are looking for.

GaIA: Generative AI for Judicial Intellectual Identity and Strategic Appellate Judgments (Subnational: Porto Velho)


-
Added by:   OECD analyst
Added on:   06 Aug 2026
Updated by:   OECD analyst
Updated on:   06 Aug 2026

undefined
GaIA is a generative AI assistant that helps appellate judges and court staff draft complex legal documents, such as case summaries and judicial decisions. It was developed to reduce judicial backlog by speeding up the writing process while preserving each judge's unique legal voice and style. This innovation ensures faster legal services for citizens while keeping humans in control of every final decision.

Name in original language

GaIA: IA Generativa para Identidade Intelectual Judiciária e Elaboração Estratégica de Acórdãos

Initiative overview

The Brazilian judiciary faces a critical backlog that directly impacts the delivery of justice to society. At the Appellate level (Second Instance), drafting complex judgments (acórdãos) requires extreme precision and must reflect the unique "intellectual voice" of each Desembargador (Appellate Judge). GaIA (AI-Assisted Judicial Cabinet) leverages this challenge into an opportunity by instituting a cutting-edge generative co-pilot model that amplifies human capability without replacing it. The core objective is the strategic drafting of reports, opinions (votes), and case summaries (ementas) to drastically boost procedural speed. 

A key innovation is Intellectual Identity Mimicry. Unlike generic AI models, GaIA is grounded in a personalised "Collection of Votes," allowing the AI to replicate the specific writing style, argumentative structure, and technical vocabulary of each judge. To enhance legal certainty and uniformity, the system features an advanced similarity search tool that locates the judge's previous rulings on identical legal issues, ensuring that new judgments remain harmonised with the chamber’s jurisprudential history. Ethical security and data governance are fundamental pillars; processing occurs within a secure local VPN infrastructure, with strict anonymisation guidelines and a contractual guarantee that judicial data is never used to train external models. 

GaIA also introduces the concept of "Specialists"—optimised and approved prompts shared within an institutional library. This ecosystem is strengthened by an integrated Peer Forum, where law clerks collaborate to refine AI interactions and share best practices. Institutionalised via Act 1629/2025, the system is evolving toward intelligent data reprocessing and detailed audit logs to ensure total accountability. GaIA solves the "robotic AI" dilemma by maintaining human critical analysis as the sovereign validator of every final judicial decision, serving as a worldwide benchmark for human-centric and ethical AI governance.

Other relevant details

Results, outcomes and impacts: Tangible results: 11,345 drafts generated (including 6,543 opinions, 3,707 reports, and 1,085 summaries) and 171 active prompts created by 53 users. There is a significant reduction in drafting time and enhanced jurisprudential standardisation. Impact is measured via productivity indicators in the PJe system. Future goals include full institutionalization and scaling to other court areas, aiming for continuous backlog reduction and improved public legal services. Challenges and lessons learned: The main challenge was ensuring that AI acted strictly as an assistant, preserving the judge's autonomy and final critical analysis. Privacy and technical inaccuracy risks were managed through governance that mandates human review and a secure network (VPN), preventing data from being used to train external models. A key lesson is that AI quality depends on the careful selection of documents used as the case's context. The success of the initiative stems from the collaborative environment in the Prompts library and the Peer Forum, where knowledge is shared to standardise high-quality decisions. This results in faster and more ethical justice for society. Additional insight: GaIA is not merely an isolated tool but a comprehensive judicial innovation ecosystem supported by robust documentation, including detailed PDF manuals and an interactive Notion version. The initiative is accompanied by a series of demonstration videos illustrating the complete case lifecycle within the system: from automatic document ingestion and classification to strategic chat interaction for refining complex judgments. A decisive factor for institutional trust is its solid legal foundation through Normative Act No. 1629/2025, which establishes rigorous ethical, security, and governance guidelines. Furthermore, real usage data—with over 11,000 drafts generated by an active user base—proves the practical utility and widespread acceptance of the tool within the senior judiciary's daily operations. The Court of Justice of Rondônia reaffirms its commitment to Human-Centric AI, ensuring that technology amplifies intellectual capacity without ever replacing the critical judgment and sensitivity of the judge. We remain fully available to the OECD to provide detailed technical demonstrations and to share our "Prompts" model with other government institutions interested in replicating this success worldwide.

About the policy initiative


Category:

  • AI policy initiatives, programmes and projects

Initiative type:

  • AI use cases/projects in the public sector

Status:

  • Active

Start Year:

  • 2024

Target Sectors:


OECD AI Principles:


Other relevant urls:


Images: