When Must Lawyers Disclose AI Use? Court Requirements for Legal Work

Lawyers must disclose AI use when a court rule or a judge's standing order requires it, and no single national rule governs legal writing. Court rules on AI disclosure in legal writing run from standing orders demanding a signed certification that names the tool, through statewide amendments such as Florida's, to districts that rely on Rule 11 and require nothing further.

The pressure behind those rules is measurable, the AI Hallucination Cases Database maintained by Damien Charlotin lists 2,149 court decisions worldwide involving AI-fabricated content, 1,473 of them in the United States, up from six worldwide in the quarter the Mata v. Avianca sanctions proceedings began. This guide covers which courts require disclosure, what a compliant certification states, whether a client must be told, and what follows when disclosure is missing.

Key takeaways

1. Whether lawyers have to disclose AI use depends on the court. Court-side rules split into two families: mandatory disclosure and certification regimes, which require an affirmative statement at filing, and responsibility-based standards, which require nothing beyond the verification duty a signature already carries.
2. Obligations attach to the individual judge far more often than to the court, so the governing rule for a matter is usually found on the assigned judge's own page rather than in the local rules.
3. Statewide rules now displace judge-by-judge variation in Florida and New York, and both replaced AI disclosure with a verification duty rather than extending disclosure.
4. Client-side disclosure runs on a separate track. Most US jurisdictions impose no blanket duty to tell a client that AI assisted a matter, and the duty arises from conditions such as fee impact, confidential inputs and the significance of the use.
5. Attorneys always remain fully responsible for AI-generated content, whatever a court rule requires.

[cta-1]

Why courts require AI disclosure in legal filings

Courts require AI disclosure in legal filings because fabricated authorities reached dockets faster than existing verification habits could catch them. These requirements emerged in response to real filing errors rather than to the technology itself, which is why the rules concentrate on accuracy and accountability rather than on how a model works.

The Mata v. Avianca case and its aftermath

Mata v. Avianca is the first widely reported AI sanctions case: Judge P. Kevin Castel imposed a $5,000 sanction on attorneys Steven Schwartz and Peter LoDuca and their firm, Levidow, Levidow & Oberman, after a brief prepared with an AI tool cited judicial opinions that did not exist. The full account of what the lawyer who used ChatGPT actually filed sits in its own guide.

AI hallucinations and the reliability problem

AI hallucinations in legal citations are fluent outputs that are factually wrong, and in legal filings they surface as fabricated authorities, inaccurate case summaries and invented procedural history. Judges addressing the problem have framed AI as a drafting and research aid whose output a lawyer is expected to check, not as a source of authority in its own right. The mechanics of why a model invents a citation, and which prompts make it more likely, are covered in a separate guide to hallucination in legal work.

Am I responsible for verifying AI output under Rule 11?

A lawyer is responsible for verifying AI output under Rule 11, and the responsibility does not depend on whether any AI disclosure rule applies. Rule 11 makes the signature on a filing a certification that the legal contentions are warranted and the factual contentions have evidentiary support, after a reasonable inquiry. Courts addressing AI-assisted filings have treated reliance on an unverified AI output as a failure of that inquiry rather than as a defense to it, and the Fifth Circuit, declining to adopt a special AI rule, reasoned that the existing certification already reaches the conduct.

Where no local AI rule applies, there is generally no obligation to state affirmatively that AI was used, and the Rule 11 verification duty still applies in full. That duty, rather than disclosure, is what courts have actually enforced against lawyers who filed AI-fabricated citations. The rule text sits with the Federal Rules of Civil Procedure.

Which federal courts have AI disclosure rules?

Federal courts with AI disclosure rules are mostly individual district judges acting through standing orders rather than whole districts acting through local rules. A survey of how federal courts are regulating generative AI in litigation, published by Husch Blackwell, describes the same pattern: a small number of mandatory certification regimes, a larger number of courts relying on Rule 11, and very little uniformity between them.

Federal court AI disclosure rules, district by district

1. Texas (Northern District): Judge Brantley Starr requires lawyers to certify, when they appear, whether generative AI was used in preparing filings, and to confirm that a human verified all statements and citations.
2. Texas (Eastern District): Rather than imposing a certification rule, the district amended its local rules to confirm that AI use does not change a lawyer's Rule 11 obligations.
3. Pennsylvania (Eastern District): Judge Michael Baylson requires affirmative disclosure of generative AI use, confirming that all citations and legal authorities have been verified.
4. New Jersey: Judge Evelyn Padin requires disclosure whenever AI is used in connection with court submissions, identifying the specific tool, describing which parts of the filing were affected, and certifying attorney review.
5. North Carolina (Western District): The Charlotte Division standing order requires certification either that no generative AI was used, with a carve-out for established legal research platforms such as Westlaw and Lexis, or that every statement and citation was verified by a human. It reaches pro se litigants too.
6. Illinois (Northern District): Judges differ. Some magistrate judges require disclosure where AI was used for legal research or drafting, while others emphasize that reliance on AI does not excuse an error under Rule 11.
7. California (Northern District): Magistrate Judge Kang requires AI-assisted documents to be identified by a notation in the title, a preliminary table or a separate concurrent notice. Judge Rita F. Lin permits AI use and emphasizes that attorneys alone bear ethical responsibility for the statements in a filing.
8. Michigan (Eastern District): The district has currently proposed, but not adopted, a local rule requiring disclosure of generative AI use in drafting, with certification that citations were verified.

Which courts declined to adopt AI disclosure rules?

Courts that declined to adopt AI disclosure rules include the Fifth Circuit, which considered a proposed certification requirement and resolved not to impose one, and the Illinois Supreme Court, which discouraged state judges from adopting blanket disclosure requirements. The reasoning in both is the same: a lawyer is already required to be accurate and candid, so a separate AI rule adds a step without adding a duty. A responsibility-based court therefore imposes no disclosure obligation while still imposing the full verification obligation.

How to find out whether a judge requires AI disclosure

Finding out whether a judge requires AI disclosure takes five checks, in this order, and the first resolves most matters:

1. The assigned judge's own page on the district court website, where standing orders, individual practices and judge-specific certification forms are published.
2. The district's local rules and general orders, which govern where an individual judge has issued nothing.
3. The state supreme court's AI policy, where the matter is in state court, because a statewide rule can preempt the orders beneath it.
4. A maintained tracker, to catch orders issued since the last check.
5. Chambers, by phone or by email, when a case is assigned and the published materials are ambiguous.

Which AI court-order trackers are worth checking

Two trackers cover step four. Law360 maintains an AI court-order tracker behind its subscription, organized by judge and updated as orders issue, which is a fast way to confirm that a judge's published order is still the current one. For teams without that subscription, the American Immigration Lawyers Association publishes a free running list of new court reporting requirements on lawyer AI use. It covers federal district and appellate courts as well as the immigration courts, and it names the judge or the court behind each requirement, which is enough to confirm whether a given matter is inside a disclosure regime.

Neither tracker explains the reasoning behind an order, and neither is a substitute for reading the order itself, which matters most where a rule is ambiguous about what counts as generative AI. The University of North Carolina law library maintains an academic survey of judicial guidance on generative AI in court that groups the orders by what they actually require.

Two situations defeat a one-time check:

1. The assigned judge has issued no order. The obligation does not disappear, because the local rules and Rule 11 still govern, and a certification offered voluntarily is sometimes the cleanest way to document a reasonable inquiry.
2. The case is reassigned mid-matter. The incoming judge's standing order governs the next filing even though nothing about the case changed, so reassignment belongs on the docket-review checklist alongside scheduling changes.

Statewide AI rules that replace judge-by-judge orders

Court rules requiring disclosure of AI use in filings are now being replaced, in two states, by statewide rules that narrowed disclosure rather than expanding it. Florida and New York each displaced the circuit-level and judge-level AI disclosure orders beneath them with a single statewide rule that attaches a verification duty to the signature on a filing instead.

Florida Rule 2.515(d)(2)

Florida Rule 2.515(d)(2) is a statewide AI certification standard for court filings, and it certifies accuracy rather than AI use. The Florida Supreme Court amended Rule of General Practice and Judicial Administration 2.515(d)(2) so that a signer represents, by filing, that the legal authorities identified in the document exist and are accurately cited. The court said it adopted the amendment principally to create a statewide, uniform replacement for the varied circuit administrative orders imposing AI disclosure and certification requirements, and an accompanying administrative order states that courts may not impose such requirements by local order, policy or practice. The amendment also carries an express sanctions provision. The Florida Bar reported the amendment and the order preempting circuit requirements when the court approved them.

New York Part 161

New York Part 161 is the Unified Court System's statewide court rule on AI use in attorney filings, and it declines to require disclosure. The policy provides that AI use in preparing court papers is permitted where it remains consistent with existing professional duties, and that attorneys should not be required to disclose AI use when submitting papers. What the policy does require is that anyone who used an AI tool review the paper and independently confirm that it contains no fabricated cases, statutes or other material. Individual courts may adopt a model certification in which signing certifies the absence of fabricated content, without any separate statement about AI.

What the trend means for a filing checklist

A statewide AI verification duty for court filings attaches to the signature rather than to a disclosure form, and it is the stable part of a filing checklist. Read together, Florida's rule and New York's policy point one way: state court systems are moving away from judge-specific AI disclosure and toward a statewide verification duty on every court filing, enforced by sanctions rather than by a form. For a practice filing in several states, the verification duty applies in every court, and the disclosure requirement has to be re-checked for each matter.

What must be disclosed about AI use in a court filing?

What must be disclosed about AI use in a court filing depends on which of the two families the court belongs to. Mandatory disclosure and certification regimes ask for an affirmative statement at filing, and they generally want the same things. Responsibility-based standards ask for none of it and still require the verification underneath.

Identifying the AI tool used

Several courts require a lawyer to name the tool rather than refer generically to AI software. Naming the specific product lets the court form a view about what kind of assistance was involved, which is why a generic reference tends to be treated as non-compliant. Where a firm runs several tools across research, drafting and review, the certification names the one that touched the filing.

Describing which sections were AI-generated

Describing which sections of a filing were AI-generated means documenting which portions were drafted or assisted by AI, and standing orders that require section-level attribution vary in the granularity they ask for from a specific range such as Sections II and III to a general description such as portions of legal research and analysis. Because the court may ask for clarification after filing, the internal record supporting the description matters as much as the description.

How should attorneys verify AI-generated citations before filing?

Attorneys should verify AI-generated citations before filing by pulling each authority in a primary source and confirming four things separately:

1. The case exists.
2. The citation is accurate, including the reporter, the court and the year.
3. The quoted language appears in the opinion.
4. The proposition the brief attributes to the case is what the opinion holds.

A fabricated authority fails the first test, but the more common defect is a real case cited for a holding it does not contain, which survives a bare existence check. Learn more about the practical mechanics of verifying an AI-assisted brief.

When and how must an AI disclosure be filed?

An AI disclosure must be filed at the time of filing in most courts that require one, and it must be conspicuous rather than buried in a footnote. Some judges accept a concurrent notice or a separate certification filed alongside the paper, and a few require a notation in the title or in a preliminary table. One timing rule is appearance-based: Judge Starr's certification is due when counsel appears, not when the first AI-assisted brief is filed, so a matter can already be out of compliance before any disclosure question arises.

Sample disclosure language and certification templates

AI disclosure certification language for a court filing runs to three or four sentences, because courts that require it focus on whether a filing clearly states AI involvement and confirms human oversight. The language below is a starting point to be adapted to the local rule, standing order or judge-specific form that actually applies.

Standard certification template

"This filing was prepared with the assistance of an artificial intelligence-based tool, [AI Tool Name]. The undersigned attorney certifies that all content, including factual statements and legal citations, was independently reviewed and verified by a human attorney prior to filing. The attorney remains fully responsible for the accuracy and substance of this submission."

Judge-specific template

"Pursuant to [Judge Name]'s standing order, the undersigned certifies that generative AI was [used / not used] in preparing this filing. If used, all AI-assisted content was reviewed, edited, and verified for accuracy by a licensed attorney prior to submission."

What an AI verification record should contain

An AI verification record should contain enough detail to show who checked what, and the minimum set is short:

1. The tool used, by product name and version where the vendor publishes one.
2. The date of use and the matter it related to.
3. The sections or tasks the tool assisted with, described at the granularity the governing order would ask for.
4. The authorities checked, and the source each one was checked against.
5. The name of the attorney who performed the verification and signed off.
6. Any output that was rejected, which is the entry that documents review rather than merely asserting it.

What a verification record cannot prove

An audit record cannot prove that an AI tool is reliable enough for court, and that limit is worth stating plainly. A verification record is evidence of a reasonable inquiry by a named human. It is not evidence that the tool is reliable, and no court has treated a vendor-generated log as a substitute for the lawyer's own verification. A verification record stays internal unless a court asks for it.

An internal policy for attorney use of legal AI platforms should include the same elements, which is what turns the record into a by-product rather than a reconstruction. A policy naming the approved platforms, the permitted tasks, the verification step and the attorney accountable for it produces the record in the ordinary course, which is more durable than assembling one after a court asks.

[cta-2]

Must lawyers disclose AI use to clients?

Most US jurisdictions impose no blanket duty for lawyers to disclose AI use to clients, and the duty arises from the circumstances of the use rather than from the technology. The analysis tracks the duty to communicate, which asks a lawyer to consult reasonably with a client about the means by which the client's objectives are pursued, and bar authorities have reached for the older guidance on temporary and contract lawyers as the closest analogy. Esquire Deposition Solutions, surveying how litigators weigh the need to disclose AI use to clients, notes that the question sharpens when the technology could affect the outcome or the expense of the representation.

Disclosure is generally indicated where one or more of these conditions is present:

1. The client asked, or the engagement terms already address AI use.
2. Confidential client information will be entered into the tool, and the lawyer is not reasonably satisfied that the provider will keep it confidential.
3. The use is significant to the representation, rather than clerical or administrative.
4. The arrangement affects fees, either because per-use costs will be billed as an expense or because the basis of the hourly charge changes.
5. The delegation is broad enough to resemble outsourcing part of the work.
6. The client's own policy or outside counsel guidelines restrict or prohibit AI use.

Where none of those conditions is present, the better reading of current guidance is that the duty to communicate does not reach the tool. Jurisdictions differ on how far they go, and that per-jurisdiction map, including the bar guidance published state by state, is a separate question from the conditions ladder above.

What a client disclosure typically says

Disclosing AI use to legal clients typically involves four things, in roughly this order:

1. The intended use, described for that matter rather than in general.
2. The benefit the client can expect, such as faster turnaround on a defined task.
3. The accuracy and confidentiality risks that remain after the firm's precautions.
4. The attorney review that stands between the output and the client.

The State Bar of Texas guidance on client communication protocols regarding AI use recommends raising the subject early, documenting consent in writing where the use is significant, and updating the client where the firm changes tools or expands the tasks AI touches.

When the fee arrangement makes disclosure unavoidable

The fee arrangement makes disclosure unavoidable whenever AI changes what the client is billed for rather than how quickly the work is done. Texas Ethics Opinion 705 concludes that a lawyer may generally bill for the actual time spent using a generative AI tool, including the time spent refining outputs and checking them, and may not bill hourly for time the tool saved. Where the lawyer pays per use, those costs may be recovered as an expense only if the client accepts that arrangement and only up to the amount actually incurred, which makes the client's acceptance a precondition for recovering those costs.

What to put in the engagement letter instead of disclosing per matter

An AI clause in a law firm engagement letter is the practical alternative to a per-matter conversation, and it is what many firms adopting AI across a practice settle on. A workable engagement letter clause covers four points:

1. The categories of tool the firm uses, named by function rather than by product.
2. That attorney review applies to every output before it reaches the client.
3. How the vendor handles client data, including retention and training.
4. The client's agreement to the arrangement, including any per-use costs.

American Bar Association (ABA) Formal Opinion 512, issued by the ABA Standing Committee on Ethics and Professional Responsibility in July 2024, points toward informed consent before confidential information is entered into a tool, which is further than some state guidance goes. The University of North Carolina law library's analysis of Formal Opinion 512 sets out that reasoning in full.

A standing clause does not close the question for every matter. Where a specific use would be a significant development in the representation, or where a client's own guidelines speak to it, the conversation happens anyway.

Ethical obligations for lawyers using AI beyond court rules

Ethical obligations for lawyers using AI beyond court rules apply whether or not a court requires disclosure, and four of the Model Rules carry most of the weight. The New Hampshire Bar Association's treatment of drafting documents with artificial intelligence works through the same rules from the bar side rather than the court side.

Model Rule 1.1: duty of competence

Competence extends to the tools a lawyer relies on, which means understanding what a system does, where it fails and how its output is checked. The test is not whether a lawyer can explain a model's architecture but whether the lawyer can state what the tool was used for and how its output was verified.

Model Rule 1.6: confidentiality obligations

Entering client information into a generative AI tool is a disclosure to a third-party service provider, and the rule permits it where the lawyer is reasonably satisfied that the confidential character of the information will be respected. Self-learning systems that incorporate user inputs into training data raise the specific risk that client information surfaces in a later response to someone else. These threshold questions settle most of the analysis:

1. Are prompts and documents retained after the session, and for how long?
2. Are inputs used to train or improve the model?
3. Who inside the provider can access the content, and under what controls?
4. Which subprocessors receive the content, and in which jurisdictions it is stored?

Model Rule 5.1: supervisory responsibilities

Partners and managers are responsible for arrangements that give reasonable assurance the firm's lawyers comply with the rules. For AI, a defensible arrangement states four things in writing:

1. Which tools are approved, and which are prohibited outright.
2. Which tasks each approved tool may be used for.
3. What the verification step is, and who performs it.
4. Who is accountable when an output reaches a client or a court unverified.

Training matters as much as the policy, because the lawyers most likely to use an unapproved tool are the ones working without supervision.

Model Rule 3.3: candor to the tribunal

A lawyer may not submit false or misleading information to a court, and where AI-assisted content turns out to contain an error, the duty to correct the record attaches promptly. Knowledge includes willful blindness, so leaving a red flag in an output unexamined does not establish a lack of knowledge.

What are the consequences of AI mistakes in court filings?

The consequences of AI mistakes in court filings run from a corrected filing to a referral to disciplinary authorities, and they escalate with how the error is handled more than with how it was made.

How is undisclosed AI use discovered?

Undisclosed AI use is usually discovered through the filing itself rather than through investigation. Opposing counsel or chambers cannot verify a citation, which surfaces the fabrication directly. Secondary routes include:

1. Document metadata, which can record the application that produced a passage.
2. Drafting patterns inconsistent with the rest of a brief, such as a citation format or a register that changes mid-argument.
3. Internal records produced in later proceedings, including the firm's own verification file.

Once AI use surfaces unexpectedly, courts tend to treat the non-disclosure as a candor question separate from the accuracy of the filing.

What sanctions have courts imposed for AI-generated filings?

Courts have imposed Rule 11 sanctions for AI-generated filings along a recognizable ladder, and a single matter can draw several rungs at once:

1. An order to correct or withdraw the filing.
2. Monetary sanctions, which in reported matters have ranged from a few thousand dollars upward.
3. Striking the offending document, or dismissing the proceeding where the defect reaches the pleading.
4. An award of the opposing party's costs and attorneys' fees.
5. Mandatory continuing legal education on the use of AI, imposed as a condition rather than as a penalty.
6. Referral to the state disciplinary authority, which is the rung with consequences outside the matter.

Florida's amended rule, published with the Rules of General Practice and Judicial Administration, now lists reprimand, contempt, striking of the document, dismissal of proceedings, costs and attorneys' fees expressly. That is a useful indication of the range a court considers available even where the local rule is silent.

AI malpractice and professional discipline risk for lawyers

AI malpractice and professional discipline risk for lawyers follows from the absence of oversight rather than from the use of a tool. A filing error that harms a client's position is a malpractice question on ordinary principles, and a bar complaint can follow the same conduct. Disclosure does not substitute for supervision, and a certification stating that an unverified output was verified creates additional exposure.

What falls due before the sanctions question resolves

Two obligations fall due before a sanctions question resolves, and both are easy to postpone while it is still live:

1. The client is entitled to be told about a material error in its own matter, under the same duty to communicate that governs the disclosure question in the first place.
2. The firm's professional liability carrier generally requires notice of a circumstance that could give rise to a claim, rather than notice of a claim already made.

Delaying either one narrows the options that remain available.

Case-specific consequences

A court that has found a fabricated citation has three responses available short of a further sanction:

1. Reading later filings in the same matter more skeptically.
2. Limiting the arguments it will entertain.
3. Requiring additional certification going forward.

These consequences outlast the sanction and reach the client's position as well as the lawyer's standing, which is the argument for raising an error before the court finds it.

Can a legal AI tool produce an AI disclosure record for a court?

A legal AI tool can produce most of the material an AI disclosure record needs, and it cannot produce the certification, because the certification is a representation by a named attorney. For transactional lawyers and in-house counsel, where no court is involved in most of the work, the useful version of this capability is a reliable account of what the tool produced and who reviewed it, kept in the ordinary course rather than reconstructed later.

Spellbook works inside Microsoft Word and Google Docs, which is where that account comes from without a separate logging step:

1. The Review feature surfaces risks in third-party paper and proposes redlines as native Word tracked changes, so each suggestion is accepted, edited or rejected by an identifiable lawyer and the document records the decision.
2. The Draft feature generates clause and agreement language grounded in a team's own playbooks and precedent rather than in open-web text, which narrows the class of output needing authority-level verification.
3. Work product stays under attorney control throughout, the platform operates under a zero data retention policy, and it holds SOC 2 Type II attestation.

The limitation is one of scope. Spellbook is a commercial-contract specialist rather than a litigation or compliance platform, so it is built for the agreements a transactional practice negotiates and not for the papers a litigator files. For a brief headed to a docket, neither the verification work nor the certification is something a drafting tool can take on, and the representation a lawyer signs remains the lawyer's.

Pricing is per seat and scales with team size, and there is a free seven-day trial that needs no sales call and no implementation project. Try Spellbook free to see what the Review feature produces on a real third-party agreement.

[cta-3]

Frequently asked questions

Do I need to disclose AI use if I used it for research, not drafting?

Sometimes. Several courts treat AI-assisted research the same as AI-assisted drafting and require disclosure where AI influenced the filing, while the Western District of North Carolina standing order carves out established legal research platforms such as Westlaw and Lexis. Where the governing order does not resolve the question, disclosing substantive AI involvement is the lower-risk course.

What if a jurisdiction has not issued any AI disclosure rules?

The verification duty still applies. Rule 11 and the state equivalents require a reasonable inquiry before signing regardless of how a document was prepared, and courts have enforced that duty in jurisdictions with no AI-specific rule at all. The practical step is to monitor the assigned judge's standing orders rather than the jurisdiction as a whole, because that is where the obligation usually appears first.

Can I use a general disclaimer instead of a case-specific certification?

Typically no. Most courts that require AI disclosure expect a certification tied to the specific filing, and a general firm policy or an engagement-letter clause does not satisfy a filing-level requirement. The two do different work: the policy governs the firm, and the certification speaks to the court about one paper.

Does a lawyer have to disclose AI use in a transactional matter with no court involved?

No court rule reaches a transactional matter, because court-side disclosure obligations attach to filings. What remains is the client-side and confidentiality analysis: whether the use is significant to the representation, whether confidential information goes into the tool, whether fees are affected, and whether the client's own guidelines restrict AI use.

Are AI-generated drafts protected by attorney-client privilege?

The leading decision answers a narrower question than the one usually asked. In a case before the Southern District of New York, Judge Rakoff held that a party's own AI-generated documents, created independently of counsel using a public tool to assess the party's legal exposure, were not protected: privilege had been waived by sharing the information with a third-party platform that is not legal counsel, and work-product protection did not attach because the materials were not prepared at counsel's direction. BakerHostetler's analysis of the ruling sets out the reasoning. Privilege still turns on confidential communication between a lawyer and a client rather than on whether a tool touched the text, so the operational lesson concerns which platforms a matter's confidential facts are allowed to reach.

Can a client require a lawyer not to use AI?

Yes, within the ordinary limits on client instructions about the means of a representation. A client may direct that AI not be used on its matters, and outside counsel guidelines increasingly say so in writing. Where a lawyer judges that the instruction makes the representation impractical, the conversation is about scope and fees rather than about whether the instruction is permissible. Whether a general-purpose assistant is an appropriate tool for legal work at all is a separate question a firm answers through its own policy.

What happens if AI-generated errors are discovered after filing?

The record must be corrected promptly. Courts expect notice, an explanation of the error and a corrected submission, and early voluntary correction is consistently treated more leniently than a correction prompted by opposing counsel or by the court. Reported sanctions cluster around matters where the lawyer defended the filing after the defect was identified.

Do disclosure requirements apply to pro se litigants?

Often yes. Several standing orders, including the Western District of North Carolina's, apply AI certification and verification requirements to pro se parties as well as to lawyers. Judges may allow latitude on form, and the accuracy requirement does not relax.

> About this content: This article is provided for informational purposes only. It does not constitute legal advice, and reading it does not create an attorney-client relationship. For legal advice on a specific situation, consult a licensed attorney in your jurisdiction. Spellbook is a software product; it supports legal professionals' work but does not replace their judgment. References to court rules, ethics opinions and AI model capabilities reflect publicly available information as of the publication date and may change.