ZLAQ
← All posts November 10, 2026

New York's New AI Ethics Guidelines: The Short Version for Practicing Attorneys

The NYSBA just approved comprehensive AI ethics guidelines. Here's what actually matters to your practice.

On October 31, 2026, the New York State Bar Association approved its Legal Ethics Guidelines for Lawyers' Use of Artificial Intelligence.

The report runs 96 numbered pages and covers everything from billing and confidentiality to AI agents that can send emails, modify records, or take actions on your behalf.

The good news is that lawyers can use AI extensively. But we remain responsible for how it's used.

Here are the practical takeaways from all 19 guidelines.

13 takeaways · tap one to open it
01You don't have to use AI, but you need to understand what you're using
Guidelines 1A, 1B: Competence, reliability and bias
  • Attorneys are not required to adopt AI.
  • If you use it, understand its capabilities, limitations and risks.
  • Evaluate reliability and potential bias, particularly when AI influences important decisions.
  • Stop relying on a tool for a particular purpose if it can't be used competently.
Takeaway: You don't need to understand how an AI model is built. You do need to understand whether it's appropriate for the work you're giving it.
02AI can assist with legal judgment, not replace it
Guidelines 2A, 3A: Client authority, diligence and verification
  • AI can assist with research, drafting, analysis and recommendations.
  • Lawyers must independently verify material output before relying on it or communicating it to others.
  • The attorney's professional judgment must support the advice.
  • Decisions reserved to the client remain the client's decisions.
Takeaway: Having Claude or Harvey draft an agreement is fine. Sending it without properly reviewing it is another matter.
03Be careful with AI meeting recorders
Guideline 3B: Recording and transcription
  • Comply with applicable recording laws, including those of other participants' jurisdictions.
  • Obtaining client consent is recommended. Other participants should be notified, and consent obtained where required.
  • Consider privilege, confidentiality and discoverability.
  • Verify AI-generated transcripts and summaries.

The guidelines specifically recommend against recording conversations unless there's a particular reason to do so.

Takeaway: Think twice before automatically inviting an AI note-taker to every client call.
04You don't always have to disclose AI use
Guideline 4A: Client communication and consent

Disclosure is required in certain circumstances, including when:

  • A client asks about AI use.
  • AI significantly informs advice on a decision reserved to the client.
  • Confidentiality obligations require informed consent.
  • AI affects the basis of fees or client expenses.

Routine administrative uses generally don't require disclosure.

Takeaway: Using AI doesn't automatically mean you need to notify every client.
05You can't bill for time AI saved you
Guideline 5A: Fees and expenses
  • Hourly billing is limited to time actually spent, including prompting, reviewing and verifying.
  • You cannot charge for hypothetical hours saved by AI.
  • General time spent learning an AI tool isn't billable to a client.
  • Flat fees don't have to decrease simply because AI makes the work faster.
  • General AI subscriptions are overhead. Reasonable matter-specific expenses may be charged when properly disclosed.
Takeaway: AI efficiency can improve profitability, especially with flat fees. It doesn't justify billing for hours you didn't work.
06Confidentiality is about where the information goes
Guidelines 6A, 6B: Confidentiality and data governance

Before using AI with protected information, understand:

  • Whether information is retained or used for training.
  • Who can access prompts, documents and outputs.
  • Whether data can cross between clients or matters.
  • What security, deletion and vendor safeguards apply.

These obligations extend to information concerning former and prospective clients.

Takeaway: Just because you're paying for an AI subscription doesn't automatically mean it's appropriate for confidential client documents. You need to understand the actual terms and how your data is handled.
07AI can create conflicts of interest
Guideline 7A: Conflicts

AI use alone doesn't create a conflict. Problems may arise when:

  • A system exposes information across client matters.
  • AI access bypasses an ethical screen.
  • An attorney has a financial, ownership or referral relationship with an AI provider.
Takeaway: If you're connecting AI to your firm's entire document library, make sure it has appropriate permissions to access each matter.
08Court filings still require independent verification
Guideline 8A: Candor to the court
  • Verify authorities, quotations and factual assertions.
  • Make sure citations actually support the propositions stated.
  • Follow applicable court-specific AI disclosure and certification requirements, including 22 NYCRR Part 161.
Takeaway: Using a specialized legal AI product doesn't relieve you of responsibility for the final filing.
09Supervising AI is part of supervising your practice
Guidelines 9A–9D: Firm, lawyer, staff and vendor supervision
  • Firms should establish appropriate policies, training and controls.
  • Supervising attorneys remain responsible for reasonable oversight.
  • Subordinate lawyers remain bound by their own ethical obligations.
  • Nonlawyers and outside vendors using AI require appropriate supervision.
Takeaway: Having an approved AI platform isn't enough. You also need to know how your employees and vendors are actually using it.
10AI agents require a different level of oversight
Guideline 9E: Agentic AI

I think this is probably the most important section for the future of law-firm automation.

There's a big difference between an AI assistant that drafts an email and one that's authorized to send it.

The guidelines call for:

  • Defining exactly what an AI system can do and what information it can access.
  • Lawyer approval before consequential or difficult-to-reverse actions.
  • Records of what the system did, when and on whose instruction.
  • Monitoring the system and suspending it when it exceeds its authority.
  • Periodic testing to make sure the system can't be manipulated by unauthorized instructions.

Examples of actions requiring approval include substantive communications, filing documents, transferring funds, changing deadlines, modifying client records and executing agreements.

Takeaway: There's a lot you can automate, but the attorney needs to remain in control of actions that have real consequences.
11Public-facing AI cannot become an unsupervised lawyer
Guideline 10A: Unauthorized practice of law
  • AI doesn't expand the jurisdictions in which an attorney may practice.
  • Public-facing systems must not mislead users into believing they're communicating with a lawyer.
  • Lawyers cannot use automated systems to provide unauthorized or unsupervised legal services.
Takeaway: A website chatbot that schedules consultations is different from one independently giving legal advice.
12Pro bono work follows the same rules
Guideline 11A: Pro bono services

AI can help expand access to legal services, but the same duties of competence, verification, confidentiality and supervision apply.

Takeaway: Just because legal services are being provided for free doesn't mean the professional obligations are any different.
13AI marketing and solicitation must be truthful
Guidelines 12A, 13A: Advertising and solicitation
  • AI-generated advertisements remain subject to attorney advertising rules.
  • Fabricated testimonials, results or endorsements are prohibited.
  • Synthetic depictions that could mislead viewers require appropriate disclosure.
  • Claims about AI capabilities must be supportable.
  • AI-assisted solicitation remains subject to the usual restrictions on contacting prospective clients.
Takeaway: AI can help create your marketing. It can't invent your track record or get around solicitation rules.

What I'd actually do this week

For attorneys already using AI, I would start with four things:

  1. Review your engagement letter. Make sure it appropriately addresses AI-related fees, necessary disclosures and recording consent.
  2. Check your AI providers' confidentiality terms. Pay particular attention to retention, training and third-party access.
  3. Review your automations. Identify which actions require your approval and whether those actions are being logged.
  4. Have a verification process. Make sure substantive AI-generated work is properly reviewed before it leaves your office.

The bottom line

I don't read these guidelines as the NYSBA discouraging attorneys from using AI. They're explaining how our existing professional responsibilities apply when we do.

To me, the most interesting part is the attention given to agentic AI. We're moving beyond AI that simply drafts and analyzes documents to systems that can actually perform tasks on our behalf.

That's where I see law-firm technology heading, and it's also where we need to be particularly thoughtful about how these systems are set up.

The point isn't to use less AI. It's to use it properly, while keeping the attorney in control.

Read the original: NYSBA Legal Ethics Guidelines for Lawyers' Use of Artificial Intelligence, Approved October 31, 2026 (PDF)

These guidelines are advisory interpretations of existing professional obligations, not amendments to the New York Rules of Professional Conduct. The report distinguishes mandatory obligations under existing law from recommended practices. This article is a condensed overview, not a substitute for the full guidance.

Want to talk through your own setup? Let's connect