# This is what Chile's official guide for using AI in the courts looks like

> Chile publishes a guide for the use of AI in courts with recommendations on ethics, transparency and data
> By Alfredo Collosa · Artificial Intelligence · 2026-06-01

Chile has published the "Basic Guide on Generative Artificial Intelligence for the Work of Judges (2026)", a document to guide the judiciary in the responsible use of generative AI tools in the daily work of the courts.

The guide acknowledges that this technology offers significant opportunities to modernize the judicial system. Among its main benefits it highlights the optimization of internal administrative processes, faster processing of cases, improved quality and predictability of judicial rulings and a reduction in access gaps between the judiciary and the public.

## Chile already has an AI guide for the courts

However, the document warns that incorporating AI also poses significant institutional and democratic risks. Among the main threats identified are the opacity of algorithmic decision-making processes, the possible dehumanization of judicial proceedings, potential breaches of due process guarantees and the progressive loss of critical cognitive and analytical skills if legal operators delegate their functions excessively to automated systems.

For this reason, the guide stresses that ongoing training for judges and officials is not an option but an ethical obligation. The goal is for those within the judicial system to fully understand the capabilities and limitations of these tools before using them at any stage of their work.

The document concludes with a call to prioritize previously validated institutional and corporate platforms to protect the sovereignty of judicial information and guarantee the strategic confidentiality of the data handled by the justice system.

## Ethical standards for the judiciary

The guide establishes that the incorporation of algorithmic systems in the judicial sphere must be subject to strict ethical criteria that prevent the violation of fundamental rights. To this end, it compiles international principles that should guide any process of technological adoption within the courts.

The text states that the implementation of AI systems cannot deepen socioeconomic inequalities or create new access barriers for people who lack digital literacy or sufficient economic resources. The guide recalls that pioneering regulations such as the European Union's Artificial Intelligence Regulation consider the use of AI in the administration of justice a high-risk activity. This classification particularly affects systems designed to assist judges in investigating facts, assessing evidence or interpreting legal and doctrinal rules.

Therefore, any deployment of these technologies requires rigorous audits and legal safeguards prior to implementation.

## Non-delegable human oversight

One of the document's central principles is that the right to be judged by a person constitutes a fundamental guarantee of due process.

AI, the guide maintains, must play exclusively a technical support role and never replace the ethical, evaluative and legal judgment of judges. The judge must at all times retain the ability to supervise, modify or reject any result generated by an algorithmic system when errors, biases or inconsistencies are detected.

According to the [document,](https://ia.academiajudicial.cl/presentacion/) the democratic legitimacy of a judicial ruling emanates from the constitutional authority of the person who issues it and not from the processing capacity of a machine. The guide establishes that any person affected by a ruling assisted by artificial intelligence has the right to clearly understand the grounds for that process.

Therefore, any draft, procedural document or evidence generated with the support of AI models must be expressly identified as such. It is also recommended that judges and officials keep detailed records of the prompts used and of their interactions with these tools to facilitate oversight, judicial control and public scrutiny.

## Algorithmic biases and discrimination

The document warns that AI systems can reproduce historical biases present in the data with which they were trained or incorporate errors derived from their design. These distortions can translate into discriminatory decisions that particularly affect minorities or vulnerable groups.

For this reason, it is up to judges to actively scrutinize the results generated by these technologies in order to preserve judicial impartiality. The guide prohibits entering confidential information or sensitive data from judicial cases into free commercial platforms.

According to the text, many providers contractually reserve the right to use the information entered by users to train future models, which could compromise the confidentiality of case files. It also emphasizes that final responsibility for any erroneous or defective ruling always rests with the signing judge, regardless of whether artificial intelligence tools were used in its preparation.

The guide also recommends creating multidisciplinary working groups made up of representatives of the Judiciary, academia and civil society to oversee the processes of acquiring and using these technologies under a "human in the loop" model.

## Chilean ethical and regulatory framework

The document highlights that the Chilean judiciary has already begun to incorporate these challenges into its regulatory framework. One of the instruments cited is Article 17 of the Judicial Code of Ethics (Act No. 244-2025), which specifically regulates the use of artificial intelligence and new technologies within the courts.

The rule prohibits automatically and unthinkingly incorporating AI-generated texts into judicial proceedings or rulings. It also establishes that automatic transcription or translation tools may only be used as preliminary support and must always be reviewed and certified by a person.

The guide also mentions the "Guide to Good Judicial Practices on Ethical Issues of the Judicial Academy", whose seventh section specifically addresses the challenges associated with artificial intelligence.

Among its main guidelines are the protection of confidential information through secure technological environments, the obligation to explain in an understandable way the use of technological tools to users of the system, the need for continuous training and the reaffirmation that technology can improve efficiency but not replace legal reasoning.

## How to write prompts for judicial work

The second part of the guide is aimed at the practical application of artificial intelligence in the daily work of the courts. The document proposes a formal structure for drafting effective prompts, made up of four essential elements:

*   **Role:** define the function the model should assume.
*   **Task:** clearly specify the requested action.
*   **Context:** provide relevant background and case data.
*   **Constraints:** set limits on format, length or style.

The guide also recommends applying principles such as maximum specificity in instructions, explicit reference to the applicable regulatory framework, precise definition of the response format, rigorous anonymization of personal data and mandatory review of any AI-generated content.

The document recalls that even the best prompts do not eliminate the risk of errors or hallucinations, so all information must be verified before being incorporated into judicial work.

## Recommended tools

The guide analyzes various artificial intelligence platforms and describes their possible applications within the judicial sphere. Among them it highlights NotebookLM, by Google, defined as one of the safest tools for analyzing case files, documents and evidentiary material because it limits its answers exclusively to the information provided by the user.

It also mentions Pinpoint, likewise developed by Google, which allows large volumes of documents, recordings and transcripts to be analyzed to identify patterns, relevant terms and actors of interest. In the field of conversational models, the guide includes ChatGPT, by OpenAI, and Claude, by Anthropic, noting that they stand out for their ability to draft complex texts, review drafts and improve the argumentative coherence of legal documents.

It also analyzes MagnarAI, focused on legal automation, and Co-Counsel, developed by Thomson Reuters, specializing in legal research and advanced document analysis.

## Practical use cases

The guide concludes with a series of video tutorials and practical examples intended to show how artificial intelligence can be used in specific tasks in the judicial sphere. These include automated transcription of hearings with speaker identification, anonymization of documents through automatic detection of sensitive data, forensic analysis of civil case files, extraction of procedural milestones and critical review of rulings to detect legal inconsistencies.

Examples of assisted drafting of judicial rulings are also included, such as the preparation of drafts related to precautionary measures in criminal proceedings. Despite these possibilities, the guide's final message is unequivocal: artificial intelligence can improve efficiency and support certain technical tasks, but responsibility, legal judgment and the final decision will remain, at all times, the exclusive competence of judges.
