What Courts Require Ai Disclosure

Over 300 federal judges now have standing orders requiring AI disclosure in court filings, and the number grows weekly. The disclosure landscape is a patchwork -- some judges require blanket certification, others only require disclosure when AI content wasn't verified by a licensed attorney, and some circuits are moving toward uniform rules.

If you're a litigator filing in federal court in 2026, you need to know every judge's AI standing order before you file. Getting this wrong isn't theoretical -- lawyers have been sanctioned for failing to disclose AI use even in courts where the requirement seemed ambiguous. Here's the complete landscape.

Federal Court AI Disclosure: The Current Landscape

As of April 2026, over 300 federal judges across every circuit have issued standing orders or local rules addressing AI use in filings. The approaches break into three categories:

Category 1: Blanket Disclosure (most restrictive). Requires attorneys to certify whether AI was used in preparing any filing, regardless of whether the output was verified. About 35% of standing orders fall here. Example: 'Counsel shall disclose in each filing whether artificial intelligence was used in drafting or researching the document.'

Category 2: Conditional Disclosure (most common). Requires disclosure only when AI-generated content was included without full human verification, or when AI was used for legal research that forms the basis of the filing. About 50% of standing orders take this approach. Example: 'If generative AI was used to draft or research this filing, counsel shall certify that all citations have been independently verified.'

Category 3: Responsibility-Based (most permissive). Doesn't require specific AI disclosure but reaffirms that counsel is responsible for the accuracy of all filings regardless of tools used. About 15% of standing orders. Example: 'Counsel remains fully responsible for the contents of any filing, including any content generated by artificial intelligence.'

The trend is clearly toward Category 1 and 2. No judge who has issued an order has moved to a less restrictive position over time.

Circuit-Level Positions

Fifth Circuit: The most structured approach. A 2024 standing order requires attorneys to certify AI use and confirm human review in all appellate filings. Several district courts within the Fifth Circuit have adopted even stricter requirements.

Second Circuit: No circuit-wide rule, but SDNY and EDNY have some of the most detailed individual judge orders in the country. Judge Castel's order (SDNY) is frequently cited as a model -- it requires disclosure of AI use and certification that human attorneys verified all citations and legal analysis.

Ninth Circuit: Running a pilot program for AI disclosure in appellate briefing as of 2025. Several district courts (CD Cal, ND Cal) have strict individual judge orders, particularly in patent and IP litigation.

Third Circuit: Individual judge orders vary widely. No circuit-wide approach yet, but the Circuit's Advisory Committee has recommended a uniform rule.

Federal Circuit: Requires disclosure for patent-related filings where AI assisted in claim construction, prior art analysis, or technical arguments. The most subject-matter-specific approach.

Seventh Circuit: Judge Easterbrook's order (the most cited) requires disclosure when AI was used for 'substantive legal analysis or research' but not for 'grammar, formatting, or other ministerial tasks.'

Other circuits: DC Circuit, First, Fourth, Sixth, Eighth, Tenth, and Eleventh all have individual judge orders but no circuit-wide rules. Expect uniform rules from most circuits by end of 2026.

State Court AI Disclosure Requirements

At least 19 states have some form of AI disclosure requirement at the trial court level:

States with formal rules or orders: - Texas: Judicial guidelines (2024) require disclosure when AI was used for drafting or research. Several district courts have additional requirements. - California: Proposed rule amendment (pending) would require disclosure in all civil filings. Many superior courts have individual judge orders already. - New York: Uniform Court System guidelines require certification that AI-generated citations have been verified. - Florida: Supreme Court standing order requires AI disclosure and human verification certification. - Illinois: Cook County Circuit Court has AI disclosure requirements for civil and criminal filings.

States with guidance but not mandatory requirements: - Colorado, New Jersey, Michigan, Ohio, Georgia, Massachusetts, and others have issued guidance encouraging disclosure but stopping short of mandating it.

States with no AI-specific rules: - The remaining states rely on existing rules of professional conduct and candor obligations. Absence of an AI-specific rule doesn't mean absence of an obligation -- Rule 3.3 (candor to the tribunal) and Rule 11/equivalent certification requirements apply to AI-generated content.

The Sanctions Record: What's Actually Happened

The case that changed everything was *Mata v. Avianca* (SDNY, 2023), where attorneys submitted a brief with AI-fabricated case citations. But the sanctions landscape has expanded significantly since then:

Publicly documented sanctions (through early 2026): - At least 8 attorneys sanctioned for AI-related misconduct in federal courts - At least 4 additional disciplinary proceedings at the state bar level - Sanctions have ranged from $5,000 monetary penalties to case dismissal to referral for disciplinary proceedings

Common patterns in sanctioned cases: 1. Attorney used consumer AI for legal research without verifying citations 2. AI fabricated cases that didn't exist or misrepresented holdings 3. Attorney filed the brief without checking 4. Opposing counsel or the court discovered the fabricated citations 5. Court found violation of Rule 11 and/or Rule 3.3

What courts are looking for: The sanctioned cases all involved failure to verify. No attorney has been sanctioned simply for using AI -- the issue is always submitting unverified AI output. Courts have explicitly said that AI use is acceptable when combined with human verification.

The Practical Compliance Playbook

Don't try to memorize 300+ standing orders. Build a system:

Step 1: Template certification language. Draft three versions of AI disclosure language -- one for blanket disclosure courts, one for conditional disclosure courts, and one for responsibility-based courts. Include all three in your filing templates.

Step 2: Judge-specific research. Before filing in any court, check the judge's standing orders for AI requirements. Sites that track this include AI Vortex's disclosure database (207+ courts), LegalTech News's tracker, and individual court websites.

Step 3: Default to disclosure. When in doubt, disclose. No attorney has ever been sanctioned for unnecessary AI disclosure. Many have been sanctioned for insufficient disclosure. The risk is entirely one-directional.

Step 4: Firm-wide tracking. Maintain a spreadsheet of every judge your firm regularly appears before and their AI disclosure requirements. Assign someone to update it monthly.

Step 5: Document your verification. Keep records of your verification process -- what tools you used, what citations you checked, what you confirmed. If challenged, you want documentation showing you took AI accuracy seriously.

The one-sentence rule: If you can't certify that every citation in your filing exists and says what you claim it says, don't file it. AI or no AI, this has always been the rule. AI just made it easier to violate.

The Bottom Line: Over 300 federal judges and 19 states have AI disclosure requirements, ranging from blanket disclosure to responsibility-based frameworks. Default to disclosing AI use in every filing -- no attorney has been sanctioned for over-disclosing, but at least 8 have been sanctioned for under-disclosing. Build template certification language and check judge-specific orders before every filing.