AI Regulation Tracker / Courts and litigation
Trinidad and Tobago Courts Adopt Binding Rules for AI in Filings
The Judiciary of Trinidad and Tobago has issued a binding Practice Direction governing the use of generative AI in court proceedings. It was Gazetted on July 9, 2026, and it requires anyone who used a Gen AI tool on a court document to certify it, while barring AI from generating the content of affidavits and witness statements.
The Judiciary of Trinidad and Tobago has put a binding rule around generative AI in litigation. The instrument is called the Practice Direction on Use of Generative Artificial Intelligence (AI). It was issued on July 7, 2026, came into effect on July 8, and was published in the Trinidad and Tobago Gazette, Volume 65, Number 118, on Thursday July 9, 2026, as Notice 1013 at page 2148. It carries the authority of the Judiciary under Chief Justice Ronnie Boodoosingh.
This is not a study or a consultation. It is a practice direction, which means it binds the way business is done in the courts. Its first paragraph sets the scope plainly: "This Practice Direction addresses the use of Generative Artificial Intelligence (Gen AI) in legal proceedings in the courts of Trinidad and Tobago. It applies to both closed-source and open-source large language models Gen AI." In other words, it does not matter whether you used a hosted commercial tool or a model you run yourself. If it is a Gen AI large language model and it touched a court document, the rule applies.
The certification requirement
The center of gravity in this practice direction is disclosure. A document that was prepared with the assistance of Gen AI has to say so, on its face, in a set form. The certificate reads: "This document was prepared with the assistance of Gen AI (state, Tool and Version) as follows: (briefly describe, in less than 100 words, how it was used)."
Read that carefully, because it asks for more than a checkbox. You name the tool. You name the version. And then you describe, in under 100 words, how you actually used it. That is a real substantive statement, not a formality. It puts the person filing the document on the record about what the machine did and what the human did.
Affidavits and witness statements
The practice direction draws a hard line at evidence. Under paragraph 7, "Gen AI must not be used to generate the content of affidavits, witness statements, or other material intended" for that evidentiary purpose. This is the part that should feel intuitive to any litigator. An affidavit is supposed to be a sworn account of what a person actually knows and did. A witness statement is a person's own account. The moment a model generates that content, you no longer have a human account, you have a plausible-sounding draft that a person signed. The court is refusing to let that substitution happen quietly.
Disclosure for research and limits on the bench
Beyond filed documents, the practice direction reaches into how AI is used along the way. Users must "disclose to the judicial officer if and when they are using Gen AI for research purposes." So it is not only the final product that gets flagged. If a Gen AI tool was part of how the legal research got done, the judicial officer is told.
The rule also turns the lens on the judiciary itself, which is worth pausing on. Judicial officers are restricted from using Gen AI in their judicial reasoning and in drafting judgments. The practice direction is explicit that "The final responsibility for the accuracy of the judgment rests with the judicial officer." A judge cannot outsource the thinking to a model and cannot hide behind one if the reasoning is wrong. The Chief Justice put the philosophy in plain terms: "We should not wish AI to be deciding cases for us. It can provide useful data for us to make human decisions."
Sanctions
This has teeth. The practice direction sets out that a court may impose the "sanction deemed necessary for the misuse of Gen AI in court proceedings, which may include striking out of the document or part of it, initiating disciplinary procedures and making adverse costs orders." Those three are not trivial. Striking a document, or part of it, can gut a filing. Disciplinary procedures reach the attorney personally. Adverse costs orders hit the wallet. A court that can do all three has a credible way to enforce the certification and the affidavit rule.
What this means for US attorneys
Let me be precise about relevance. This is a Trinidad and Tobago court rule. It binds practitioners, litigants, and judicial officers in Trinidad and Tobago. It has no legal force over a filing in a US court, and nothing about your obligations in the Southern District of anywhere changed because of it.
The reason to read it anyway is comparative, and it is a strong signal. Plenty of US federal and state judges have issued standing orders about AI over the last couple of years, and many of them ask lawyers to certify whether AI was used and to confirm that citations were checked by a human. What Trinidad and Tobago has done here goes a step further in two ways. First, the certificate is not a yes-or-no attestation, it asks you to name the specific tool and version and to describe the use in a short narrative. Second, it pairs that with an independent-verification posture and an explicit bar on AI-generated evidence, and it applies across the whole justice system rather than one judge's courtroom. That is a more structured, jurisdiction-wide model than most US standing orders, and it is the kind of approach a US court or bar could borrow.
If you practice in a jurisdiction that has an AI standing order, the practical thread is the same one this practice direction makes formal. Know which tool and version you used, be ready to describe how you used it, never let a model write sworn testimony, and keep a human accountable for every citation and every judgment call. The court in Port of Spain has just made that a rule with sanctions attached. Wherever you practice, it is good discipline regardless.
Questions professionals are asking
Is the Trinidad and Tobago AI practice direction binding?
Yes. It is a Practice Direction issued by the Judiciary of Trinidad and Tobago, effective July 8, 2026 and Gazetted July 9, 2026. It binds the conduct of proceedings in the courts of Trinidad and Tobago. It is a court rule, not research or guidance.
What does the certification requirement say?
A document prepared with the assistance of Gen AI must carry a certificate stating: "This document was prepared with the assistance of Gen AI (state, Tool and Version) as follows: (briefly describe, in less than 100 words, how it was used)." You name the tool and version and describe how you used it in under 100 words.
Can Gen AI be used to write affidavits or witness statements?
No. The practice direction states that Gen AI must not be used to generate the content of affidavits, witness statements, or other material intended for that purpose. Users must also disclose to the judicial officer if and when they are using Gen AI for research.
Does this affect US attorneys or US court filings?
Not directly. It binds only the courts of Trinidad and Tobago and has no force over US filings. For US practitioners it is a useful comparative benchmark, a Commonwealth court requiring mandatory tool certification and independent verification, going further than most US standing orders.
What happens if someone misuses Gen AI in a filing?
The court may impose the sanction it deems necessary, which may include striking out the document or part of it, initiating disciplinary procedures, and making adverse costs orders.
RELATED BRIEFINGS
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any rule or practice direction applies to your situation with qualified professionals in the relevant jurisdiction.