Ethics matrix
Where the bar stands on AI assistance.
A condensed, lawyer-readable map of the published guidance that matters most for an in-hearing research tool: the ABA, NJ, NY, CA, TX, and FL. Each entry lists the controlling authority, the date, the key requirements, and what Local Motion's posture means under that authority.
Use this as a pointer, not a substitute. Rules and opinions evolve; your jurisdiction may have a standing order or local rule this page doesn't reach. Always verify against your bar's current publications and your court's specific orders. This page is informational and is not legal advice.
The pattern across jurisdictions
Bar guidance converges on the same four duties: competence (you have to understand the tool), confidentiality (you have to know where your client's information goes), candor (you have to be straight with the court about what you used), and supervision (the work is still yours). Every jurisdiction below frames the analysis under those four headings, even when they use different rule numbers.
Local Motion is built so each of those duties is easier to discharge than to violate — the citation-lock supports competence, the bring-your-own-key + Managed AI options support confidentiality, the prewritten templates support candor, and the review-and-approve library workflow supports supervision.
American Bar Association
ABA Formal Opinion 512 — "Generative AI Tools"
What it says
- Competence (Rule 1.1). Lawyers must develop a reasonable understanding of any generative AI tool used, including its known limitations and the risk of inaccurate output. "Blind reliance" is not competence.
- Confidentiality (Rule 1.6). A lawyer must obtain informed client consent before inputting client information into a self-learning GAI tool. Lawyers must evaluate the tool's data-handling, retention, and training practices.
- Communication (Rule 1.4). Lawyers should communicate with clients about meaningful use of GAI in the matter where the use is significant to the representation.
- Candor & supervision (Rules 3.3, 5.1, 5.3). Lawyers remain responsible for the accuracy of work product GAI helped produce. Citations and factual assertions must be independently verified.
- Fees (Rule 1.5). Time saved by GAI may not be billed as if the lawyer spent it; charges to the client must reflect actual work performed.
What Local Motion's posture means under Opinion 512
- Competence. The citation-lock structurally narrows the AI's output to your reviewed library — limiting the inaccuracy surface to misapplication of real cases, which a competent lawyer is already trained to catch.
- Confidentiality. Audio stays on your machine; only transcribed text is sent to the LLM. Managed AI runs on isolated infrastructure with no training-on-customer-data posture. BYO keys lets you use a provider you've already evaluated.
- Communication. Use of a research-retrieval tool is generally not "significant" to a representation in the way drafting pleadings via AI would be. Disclosure templates available if the matter or client expects it.
- Verification. Stated plainly in our terms and in every disclosure template: the lawyer is responsible for independent verification of any authority cited.
Search "ABA Formal Opinion 512" on americanbar.org for the current published text and any subsequent guidance from the ABA Standing Committee.
New Jersey
NJ Supreme Court — Preliminary Guidelines on Use of AI by Lawyers
What it says
- NJ does not require advance court approval for AI use; existing Rules of Professional Conduct (RPCs) govern.
- Accuracy: lawyers remain personally accountable under
RPC 3.3(candor) for the truthfulness of statements and citations submitted to a tribunal. Use of AI does not transfer that duty. - Confidentiality:
RPC 1.6applies to AI use. Lawyers must understand whether client information will be retained or used to train the model. - Misconduct:
RPC 8.4prohibits dishonesty regardless of the technology used; AI-generated misstatements are still misconduct attributable to the lawyer. - The Court reserved the right to adopt formal rules after further study; lawyers should monitor for updates.
Practice notes
- No NJ rule currently mandates disclosure of AI use, but several federal judges sitting in NJ have issued individual standing orders requiring it for filings.
- Family Part practitioners: there is no NJ-specific prohibition on a litigant (including a pro se litigant) using a research tool during a remote hearing.
- Two-party-consent for recording: NJ is a one-party-consent state for audio recording, but live transcription used as a research aid (not stored or transmitted as audio to third parties) is not recording in the wiretap sense.
NJ Courts publishes notices to the bar on njcourts.gov. Search "NJ Supreme Court AI guidelines 2024" for the current version, and check the Advisory Committee on Professional Ethics (ACPE) for any subsequent formal opinion.
New York
NYSBA Report on AI; NYC Bar Formal Opinion 2024-5
What they say
- Lawyers using GAI must understand its capabilities and limitations sufficient to fulfill their duty of competence under
NY RPC 1.1. - Confidentiality under
NY RPC 1.6applies to any system that receives client information. Lawyers must vet tools' data-handling. - Supervision: under
RPC 5.1and5.3, lawyers remain responsible for AI-produced work product just as for that of a non-lawyer assistant. - Candor under
RPC 3.3: AI-produced citations and factual assertions must be verified before being submitted to a court. - Billing: AI time savings should be reflected in fees; "ghost" charging of AI-saved time as if it were lawyer time is impermissible.
Federal-court overlay (SDNY, EDNY)
- Several individual judges have issued standing orders requiring counsel to certify whether generative AI was used in preparing filings and, if so, that a human verified each citation.
- The 2023 sanctions order in Mata v. Avianca (S.D.N.Y. 2023) — counsel cited cases ChatGPT fabricated — set the practical floor for AI-related misconduct nationwide.
nysba.org for the NYSBA AI Task Force Report. nycbar.org for Formal Opinion 2024-5. Check your assigned judge's individual rules in SDNY/EDNY for standing orders on AI disclosure.
California
State Bar of California — Practical Guidance for Generative AI
What it says
- Competence under
CA Rule 1.1requires understanding the tool's known limitations, including the risk of inaccurate or fabricated output. - Confidentiality under
CA Rule 1.6andBus. & Prof. Code § 6068(e): prompts containing client information may not be input into a GAI tool that retains, learns from, or shares that data without informed client consent. - Output verification: lawyers must review and confirm the accuracy of AI output before relying on it.
- Communication under
CA Rule 1.4: significant AI use should be disclosed to the client; routine use of well-vetted tools generally need not be. - Charges to the client under
Rule 1.5: bill for the lawyer's time and judgment, not for AI computation.
Practice notes
- CA was the first major bar to issue practical guidance — many other states cite COPRAC's framing.
- CA is a two-party-consent state for recording (
Penal Code § 632). Live transcription used as a research aid is materially different from recording, but lawyers should still be mindful of how Zoom/Teams' recording disclosures interact with any client-confidential audio.
calbar.ca.gov for the current COPRAC practical guidance. Search "California Practical Guidance for the Use of Generative AI."
Texas
State Bar of Texas — AI Task Force Report
What it says
- Recommends CLE and education on AI capabilities and limitations for competence under
Texas Disciplinary Rule 1.01. - Recommends explicit confidentiality vetting under
Rule 1.05before inputting client data into any third-party AI tool. - Recommends that lawyers verify all AI-generated citations and factual assertions before submission, with candor obligations under
Rule 3.03unchanged.
Federal-court overlay (N.D. Tex.)
- Judge Brantley Starr (N.D. Tex.) was among the first federal judges to require an AI disclosure certification with any filing — the "Mandatory Certification Regarding Generative Artificial Intelligence" — requiring counsel to certify either no AI use or human verification of every cite.
- Several other N.D. Tex. and W.D. Tex. judges have followed with similar standing orders.
texasbar.com for the Taskforce report. For federal practice, check the assigned judge's individual rules — standing orders on AI vary by chambers.
Florida
Florida Bar Ethics Opinion 24-1 — Generative AI
What it says
- Confidentiality (Florida
Rule 4-1.6). A lawyer must obtain a client's informed consent before disclosing confidential information to a third-party AI provider; the lawyer must also evaluate the provider's data-protection posture. - Supervision (
Rule 4-5.3). Lawyers must oversee AI-assisted work to the same standard as work performed by non-lawyer assistants. - Fees (
Rule 4-1.5). Charges must reflect actual work performed; AI cost may be a reasonable disbursement if disclosed and not double-billed. - Advertising (
Rule 4-7). Lawyers may not describe their use of AI in ways that mislead a prospective client about the lawyer's services or qualifications.
Practice notes
- FL Opinion 24-1 explicitly notes that using AI is not per se a violation of any Florida rule — the issue is how the lawyer uses it. Local Motion's posture (lawyer-curated library; no fabricated output; lawyer remains responsible) is the kind of use the opinion contemplates as acceptable.
floridabar.org for Opinion 24-1 and any subsequent guidance from the Professional Ethics Committee.
Federal courts (cross-jurisdictional)
Judge-by-judge standing orders on AI disclosure
At the federal level there is no uniform rule, and there will not be one soon. As of mid-2026 dozens of district judges across the country have issued individual standing orders, often requiring one of the following:
- A signed certification that either no AI was used in preparing a filing, or that a human verified every citation generated with AI help.
- A footnote in the filing identifying any portion drafted with AI assistance.
- An obligation to flag, on the record, any AI tool used during oral argument.
The first place to check is your assigned judge's individual rules. Most chambers publish these on the court's website. If your judge has not yet issued an order, the safe default is to follow the strictest one your circuit's bench has adopted.
uscourts.gov and the individual chambers pages of your assigned judge. The American Bar Association's Standing Orders on AI tracker is also a useful informal aggregator.
How this page is maintained
We review this matrix quarterly and re-publish whenever a covered jurisdiction issues a formal opinion, a state supreme court adopts a rule, or a federal judge in a covered district issues a notable standing order. The disclosure templates shipped with paid plans are updated on the same cadence.
If a published opinion or rule we missed is relevant to your practice, please tell us. We do not edit history — versioned snapshots are kept so a customer can document the guidance that was current at the time of any past hearing.
Have a specific objection rather than a general rules question? The objections page answers the questions partners actually raise — Mata v. Avianca, work-product privilege, the third-party-vendor exception, Rule 1.1 competence, Rule 3.3 candor, Rule 5.3 supervision, Rule 5.5 / UPL for pro se, stale law, sanctions exposure — in the vocabulary of the bar.
Disclaimer. This page summarizes publicly available bar guidance for general orientation. It is not legal advice, does not establish an attorney–client relationship, and should not substitute for review of current primary sources and the rules of your specific jurisdiction and court. Where this page describes Local Motion's product behavior, that description is intended to be accurate as of the publication date; product behavior may evolve.