guidance diffAI in CourtsAI Governance
AI guidance beyond the big jurisdictions: India, Brazil, Hong Kong, the Gulf
AI guidance in India (Kerala), Brazil (CNJ Res. 615), Hong Kong, and the Gulf (QFC) — judicial and regulator documents beyond the big jurisdictions.
Beyond the jurisdictions with the deepest case records, courts and regulators in India, Brazil, Hong Kong, and the Gulf have issued their own AI guidance. This page collects the verified documents, with each entry’s verification status noted, so the picture is honest about what has been read at primary and what has only been confirmed at the issuing body.
India — Kerala High Court
The Kerala High Court issued a “Policy Regarding Use of Artificial Intelligence Tools in District Judiciary” on 19 July 2025, binding on all district courts in the state. It is the first formally documented binding Indian judicial AI policy.
Its position is assistive-only and strict: AI tools “shall not be used to arrive at any findings, reliefs, order or judgment under any circumstances”; cloud AI use is restricted to a whitelist of tools formally approved by the High Court or Supreme Court; all outputs, especially citations, must be verified; courts must keep an audit log of every AI use; and violations may bring disciplinary action. It sits above the Indian incident record, where tribunals themselves relied on fabricated citations. (Verified: primary document read.)
Brazil — CNJ Resolution 615/2025
Brazil’s National Council of Justice issued Resolution No. 615 on 11 March 2025, a risk-tiered governance framework for AI in the judiciary that supersedes the earlier Resolution 332/2020.
It classifies AI by risk, requires effective and periodic human oversight of any AI that could influence adjudication, and prohibits “excessive risk” uses: systems that preclude human review, predictive profiling to forecast crime or recidivism, and the use of external generative AI for high-risk purposes. All AI solutions must be catalogued in the CNJ’s “Sinapses” platform. Separately, the São Paulo bar’s ethics tribunal (OAB-SP) approved a parecer on 16 April 2026 requiring human analysis before any court submission and partner supervision of AI use. (CNJ Resolution: content confirmed at the official record and multiple legal analyses; the CNJ primary site geo-blocks this environment, so the full text is flagged for a direct read before publication.)
Hong Kong — Judiciary and Law Society
The Hong Kong Judiciary issued “Guidelines on the Use of Generative Artificial Intelligence” for judges, judicial officers, and support staff in July 2024. Judicial decisions must be made independently and personally; generative AI must never usurp the judicial function; all output must be verified, with an explicit warning about fictitious cases; and no confidential material may be entered into public chatbots. (Verified: primary document read.)
The Law Society of Hong Kong published a position paper on AI’s impact on the profession in January 2024. A December 2025 item often cited as a “circular” is, accurately, a Professional Indemnity Scheme risk bulletin on generative AI in practice, hosted on the Law Society site rather than a binding practice rule. (Bulletin read at primary; the 2024 position paper confirmed via secondary sources only.)
Gulf — Qatar Financial Centre
The Qatar International Court and Dispute Resolution Centre issued Practice Direction No. 1 of 2026, with accompanying Practice Guidance, on 6 January 2026, governing litigants and their representatives before the QFC Court and Regulatory Tribunal. It requires independent verification of all submissions, prohibits entering confidential or privileged information into public AI tools, does not mandate prior disclosure of AI use unless the court asks, and warns of disciplinary sanctions for breach. It was issued in direct response to Sheppard v Jillion, where a lawyer was found in contempt over fake AI-generated cases. (Title, date, and requirements confirmed on the QICDRC’s own site; the full practice-direction PDF was not separately available and is flagged for a direct read.) No dedicated DIFC Courts AI practice direction was found.