Key takeaways
- NYC Bar Formal Opinion 2024-5 gives general guidance on generative AI and calls the state bar's suggestion of new AI rules premature[1].
- A lawyer using an AI chatbot for intake must supervise it adequately, because it may fail to say it is not a lawyer or appear to give advice[1].
- Lawyers may bill actual time spent with AI but not the time the work would have taken without it[1].
- NYC Bar Opinion 2026-2 says to get every other party's consent before an AI tool records a call, and to default to not recording prospective clients[2].
Who has spoken in New York
New York lawyers get their AI guidance mainly from bar association ethics committees and reports rather than from a statewide rule of conduct. The New York State Bar Association's task force issued a report in 2024 that suggested adopting some AI-specific rules. The New York City Bar Association's Committee on Professional Ethics followed with Formal Opinion 2024-5 in August 2024, offering general guidance and describing the state bar's call for new rules as premature given the pace of change[1]. The City Bar has since added opinions on AI tools that record, transcribe and summarize conversations: 2025-6 for conversations with clients and 2026-2 for conversations with people who are not clients[2].
The audience is large. New York had 11,745 offices of lawyers in 2023[3], and its law offices employed 114,660 people, second only to California[4].
Opinion 2024-5: the core guidance
Opinion 2024-5 runs through the New York Rules of Professional Conduct one duty at a time: confidentiality, conflicts, competence and diligence, advertising and solicitation, compliance with law, supervision, communication, fees and candor[1]. Several points are directly useful for firm policies.
- Confidentiality: a lawyer must not input confidential client information into any generative AI system that lacks adequate confidentiality and security protections.
- Judgment: professional judgment cannot be delegated to AI and remains the lawyer's responsibility.
- Review: a lawyer must review all AI output, including analysis and citations, before using it for client purposes or submitting it to a court.
- Marketing: lawyers must not use AI to make false statements, to send solicitations the rules would otherwise prohibit, or to pose as a real person communicating with prospective clients.
Intake chatbots need real supervision
The opinion addresses intake directly. A lawyer using a generative AI chatbot for client intake must adequately supervise it, and a high degree of supervision may be required where ethical problems are likely. Its examples: a chatbot may fail to disclose that it is not a lawyer, or may attempt or appear to give legal advice, increasing the risk that a prospective-client or lawyer-client relationship is created[1].
| Risk named in 2024-5 | Control |
|---|---|
| Bot does not say it is not a lawyer | Fixed opening disclosure that cannot be edited by the model |
| Bot appears to give legal advice | Blocked topics and a forced handoff to a lawyer for legal questions |
| Unintended relationship forms | Minimal fact collection, early conflict check, clear statement that the firm has not agreed to act |
| Bot poses as a real person | No human names or photos; the assistant is described as AI |
The same reasoning fits AI on the phone. If an SI Employee such as Ava answers your New York line, the greeting your firm approves should say it is the firm's AI assistant and not a lawyer, it should follow only the intake questions the firm signed off, and legal questions should go to a lawyer.
Telling clients and billing
On communication, the opinion says a lawyer should consider disclosing the intent to use generative AI that is not generally understood to be routinely used by lawyers, particularly when explaining fees and disbursements[1]. Clients often have no idea either way: 71% of law firm clients surveyed by Thomson Reuters did not know whether their firms used generative AI tools[5].
On fees, a lawyer may use AI to work more efficiently and charge for actual time spent, such as writing and refining prompts or reviewing and editing output, but must not charge hourly fees for the time that would have been spent without AI[1]. Write that into your fee agreement so clients see it before the first bill.
Beyond ethics rules: the law around the tool
Opinion 2024-5 reminds lawyers that AI raises legal issues outside the professional conduct rules, including AI-specific laws, privacy laws, cross-border data transfer laws, intellectual property and cybersecurity, and that a lawyer must comply with the law and cannot assist a client in conduct the lawyer knows is unlawful[1]. For a firm, that turns into a short legal review of each tool alongside the ethics review.
- Where is the data processed and stored, and does any client or matter restrict data leaving the country?
- Does the vendor's contract give it rights to reuse inputs or outputs?
- What security controls and breach notification commitments does the vendor offer?
- Do any client engagement terms or outside counsel guidelines limit AI use on their matters?
Conflicts and shared tools
Conflicts appear in the opinion's list of duties for a reason. A tool that retains inputs across matters can, in principle, surface one client's information while the firm works for another, including an adverse party. Prefer tools that separate data by matter or do not retain inputs at all, restrict who can query collections that hold several clients' documents, and include AI tools in the information barriers the firm uses when a lateral hire or a screened lawyer is involved.
The same concern drives the City Bar's caution about recording prospective clients: information captured by a tool is information the firm holds, with all the conflict consequences that follow.
Recording: stricter than the criminal law
New York's eavesdropping law lets a party to a conversation record it, which makes New York a one-party consent state for criminal purposes. The City Bar's ethics guidance asks for more. Opinion 2026-2 concludes that, as with clients, an attorney should obtain the consent of all other parties to a call before recording it with an AI tool, and should consider whether recording, transcribing and summarizing is tactically wise in the circumstances[2].
For prospective clients, the opinion describes a paradox. A record can help the firm decide whether to accept a case, document why it declined and show that only minimal confidential information was obtained. But a recording accessible inside a firm that turned the person down creates a risk of information sharing and conflicts. Its suggested default, absent a good reason in a particular case, is not to record the conversation[2].
Firms that record intake calls for quality and accuracy should therefore make a documented decision: why the firm records, how consent is obtained from everyone on the call, who can access recordings of declined inquiries and when they are deleted.
How New York's approach differs
| Topic | New York City Bar | Elsewhere |
|---|---|---|
| New AI-specific rules | Premature; apply existing rules | California and Florida also apply existing rules through guidance |
| Intake chatbots | High supervision; must not appear to be a lawyer or give advice | Florida requires an AI and not-a-lawyer disclaimer |
| Disclosure to clients | Consider disclosing non-routine AI use, especially with fees | ABA 512 ties disclosure to use and impact |
| Billing | Actual time; not the time it would have taken | Same direction in ABA 512, California and Texas |
| Recording with AI tools | Consent of all parties; default not to record prospective clients | Criminal law in many states, including New York, requires only one party |
Where the profession stands
Attitudes inside firms are mixed but leaning positive. 59% of law firm professionals and 57% of corporate legal departments believe generative AI should be applied to their work[6], and half of law firm professionals are excited or hopeful about it, while 10% see it as a significant threat to revenue[7]. Guidance like New York's is what turns that interest into safe practice.
Source: Thomson Reuters Institute, 2025[6]
Show the data
| Item | Value |
|---|---|
| Law firm professionals | 59% |
| Corporate legal departments | 57% |
Example: a Brooklyn firm's intake review
Example: a fictional Brooklyn immigration practice uses an AI tool to answer evening calls and a meeting assistant to record consultations. After reading Opinions 2024-5 and 2026-2, it changes three things. The phone greeting now says the assistant is AI and not a lawyer and asks permission to record. Consultation recordings start only after everyone present agrees on the record. Recordings of declined inquiries are deleted after the conflicts entry is made, and the firm keeps only a short written note of why it declined.
Steps for New York firms
Adopt a written AI policy
Cover approved tools, confidentiality checks, mandatory review of output and billing.
Fix intake disclosures
Every AI that speaks to the public says it is AI and not a lawyer, and never uses a human persona.
Decide on recording
Record only with a documented reason, obtain everyone's consent and set access and deletion rules for prospective-client recordings.
Update engagement letters
Explain AI use that clients would not assume and how it affects fees.
Review regularly
New York's guidance has been updated more than once since 2024; schedule a policy review at least yearly.
For neighboring states, see New Jersey and Pennsylvania guidance; for client conversations about AI, see disclosing AI use to clients and prospects.
Frequently asked questions
Has New York adopted rules on lawyers' use of AI?
Guidance comes mainly from bar association opinions and reports, chiefly NYC Bar Formal Opinion 2024-5, which applies the existing Rules of Professional Conduct to generative AI. Courts and individual judges may also set their own filing requirements, so check local rules.
Can New York lawyers use AI chatbots for intake?
Yes, with adequate supervision. Opinion 2024-5 warns that a chatbot may fail to disclose it is not a lawyer or appear to give legal advice, and that a high degree of supervision may be needed.
Do New York lawyers need everyone's consent to record calls with AI?
New York's criminal law allows one-party consent, but NYC Bar Opinion 2026-2 says attorneys should obtain the consent of all other parties before recording a call with an AI tool.
How should New York lawyers bill AI-assisted work?
For actual time spent, including prompting and reviewing output, not the time the work would have taken without AI.
References
New York City Bar Association, Committee on Professional Ethics (2024). Formal Opinion 2024-5: Ethical Obligations of Lawyers and Law Firms Relating to the Use of Generative Artificial Intelligence in the Practice of Law (opens in a new tab) โ
New York City Bar Association, Committee on Professional Ethics (2026). Formal Opinion 2026-2: Ethical Issues Affecting Use of AI to Record, Transcribe, and Summarize Conversations with Persons Who Are Not Clients (opens in a new tab) โ
U.S. Census Bureau (2023). County Business Patterns 2023: state file (cbp23st) (opens in a new tab) โ
U.S. Census Bureau (2023). County Business Patterns 2023: state file (cbp23st) (opens in a new tab) โ
Thomson Reuters Institute (2025). 2025 GenAI report: Executive summary for legal professionals (opens in a new tab) โ
Thomson Reuters Institute (2025). 2025 GenAI report: Executive summary for legal professionals (opens in a new tab) โ
Thomson Reuters (2025). Generative AI Adoption Nearly Doubles as Professional Services Reach Crossroads (opens in a new tab) โ