What Is Ai Disclosure Order

An AI disclosure order is a standing order or local rule issued by a judge requiring attorneys to certify whether artificial intelligence was used in preparing court filings. These orders typically mandate that lawyers disclose which AI tools were used, confirm that all citations have been verified by a human attorney, and accept full responsibility for the accuracy of the filing.

As of early 2026, over 300 federal and state judges have issued some form of AI disclosure requirement. This isn't a trend — it's the new baseline. The compliance landscape is fragmented (every judge's order is slightly different), but the direction is uniform: courts want to know when AI touched a filing, and they want lawyers on the hook for the result.

What AI Disclosure Orders Require

Most AI disclosure orders share four common elements, though the specifics vary by judge. 1. Disclosure of AI use: attorneys must state whether AI was used in researching, drafting, or editing the filing. Some orders define AI broadly (including grammar tools), while most target generative AI specifically. 2. Citation verification: the attorney must certify that all legal citations — cases, statutes, regulations — have been independently verified against primary sources. This is the direct response to hallucination cases. 3. Accuracy certification: the attorney accepts personal responsibility for the accuracy of all factual statements and legal arguments, regardless of whether AI generated them. 4. Tool identification: some orders require identifying the specific AI tool used (e.g., "Claude 3.5 Sonnet" or "CoCounsel"). Others just require a general disclosure. The variation between judges is the compliance headache. What counts as "AI use" in one courtroom may not trigger disclosure in another.

Who Issues Them and Where

AI disclosure orders come from three sources. Individual judges: the majority of orders are standing orders from specific judges, typically in federal district courts. Judge Brantley Starr (N.D. Texas) issued one of the first and most cited orders. Local court rules: some courts have adopted AI disclosure as part of their local rules, applying to all judges in the district. The Eastern District of Texas adopted court-wide AI rules in 2024. State courts: several state court systems (Texas, California, New Jersey, Florida) have issued guidance or rules on AI disclosure at the state level. The Federal Judicial Center has published model language for AI disclosure orders, which is accelerating adoption. Geographic concentration is heaviest in Texas, New York, California, and Illinois — but orders exist in virtually every federal circuit. Check the specific judge's standing orders for every case, every time.

Penalties for Non-Compliance

The penalty structure has escalated rapidly since the first AI sanctions in 2023. Monetary sanctions: fines range from $5,000 to $109,700 in documented cases. The $109,700 figure comes from a 2024 case where attorneys submitted AI-generated filings with fabricated citations. Case consequences: courts have dismissed motions, struck filings, and in extreme cases impacted case outcomes due to AI-related misconduct. Professional consequences: bar referrals for ethics violations, CLE requirements (courts ordering attorneys to complete AI-specific training), and reputational damage documented in published opinions. The escalation pattern is clear: early cases (2023) resulted in warnings and modest fines. By 2025-2026, courts are treating AI-related failures as seriously as any other form of misrepresentation to the court. The Mata v. Avianca attorneys were sanctioned and reported to the bar — and their case is now cited in virtually every AI disclosure order.

The Compliance Landscape in 2026

Managing compliance across 300+ different disclosure requirements is the practical challenge. No national standard exists — each order has its own scope, definitions, and requirements. A firm litigating in 15 federal districts may face 15 different disclosure frameworks. Best practices for managing this: Maintain a jurisdiction tracker — a database of AI disclosure requirements for every court where the firm has active matters (AI Vortex maintains the most comprehensive public tracker of AI disclosure orders by federal district). Default to disclosure — when in doubt about whether a judge requires it, disclose anyway. No court has ever sanctioned an attorney for unnecessary AI disclosure. Standardize internal certification — create a firm-wide AI certification template that attorneys complete before filing any AI-assisted work product.

What Triggers Disclosure (and What Doesn't)

The gray areas cause the most anxiety. Clearly triggers disclosure: using ChatGPT, Claude, Harvey, or CoCounsel to research case law, draft sections of a brief, or generate legal arguments. Clearly doesn't trigger disclosure: spell-check, grammar tools (Grammarly), basic word processing features, and standard Westlaw/LexisNexis keyword searches. Gray area: using AI to "polish" or "reorganize" existing attorney-written text, using AI for initial brainstorming but writing the final product from scratch, using AI-enhanced features within traditional legal research platforms. Most practitioners adopt the "reasonable attorney" standard: if a reasonable attorney would consider the tool's contribution material to the filing, disclose it. This errs on the side of transparency, which is exactly where courts want practitioners to be.

The Bottom Line: AI disclosure orders require attorneys to certify whether AI was used in court filings, verify all citations, and accept responsibility for accuracy. Over 300 judges have issued them, with sanctions ranging from $5,000 to $109,700 for non-compliance. There's no national standard — every judge's order is different. Default to disclosure when in doubt. No attorney has ever been sanctioned for disclosing too much.