NYC Bar Association Formal Opinion on AI Tools: What Attorneys Need to Know

Opinion Overview

The New York City Bar Association's Committee on Professional Ethics has issued formal guidance addressing how attorneys may — and must — approach the use of AI tools in legal practice. The NYC Bar's opinion builds on the existing New York Rules of Professional Conduct (NY RPC), applying established rules to AI-specific scenarios rather than creating new obligations from scratch.

The opinion is notable for its scope: it addresses generative AI tools broadly, including large language model-based research and drafting assistants, rather than limiting its analysis to any single product category. It also engages directly with the practical questions practitioners are actually asking — whether client data can be entered into AI systems, who bears responsibility for AI-generated errors, and what supervision of AI output looks like in practice.

Structured Record

Core Obligations the Opinion Addresses

Competence (NY RPC 1.1)

The opinion applies the competence standard to AI use in a way that tracks the ABA's Model Rule 1.1 Comment 8 approach: attorneys must understand the technology they deploy well enough to use it responsibly. This does not mean attorneys need to understand transformer architecture. It does mean they need to understand what a given tool does and does not do — its scope, its known failure modes, and the conditions under which its output should not be trusted without independent verification.

Practically, the opinion signals that an attorney who submits AI-generated legal research to a court without independently verifying citations has likely failed the competence standard, regardless of whether the tool was marketed as a legal research product. The hallucination problem is specifically acknowledged — the opinion notes that generative AI systems can produce plausible-sounding but fabricated citations, and that attorneys cannot delegate the verification obligation to the tool itself.

Confidentiality (NY RPC 1.6)

This is where the opinion does the most practical work. It identifies a clear threshold question attorneys must answer before entering client information into any AI system: does the vendor's data handling create a risk of disclosure prohibited by NY RPC 1.6?

The opinion frames the analysis around three variables: whether client data is used to train the model, whether it is retained by the vendor, and whether it could be accessed by third parties. Consumer-tier AI products — tools not specifically configured for legal use, without enterprise data agreements — present the highest risk under this framework. The opinion does not categorically prohibit their use, but it requires attorneys to conduct a reasonable inquiry into data handling before using them for client matters.

Supervision (NY RPC 5.1 and 5.3)

The supervision analysis is the opinion's most consequential section for law firm management. NY RPC 5.1 governs supervisory responsibilities between attorneys; 5.3 covers supervision of non-attorney personnel. The opinion treats AI tools as falling within the 5.3 framework — the attorney who deploys an AI tool bears the same supervisory responsibility for its output as they would for work product prepared by a paralegal or junior associate.

This has direct implications for billing and workflow design. If a partner assigns AI-assisted research to an associate who then submits it without review, the partner's supervisory obligations under 5.1 are engaged. Firm-level policies on AI use — who can use which tools, what review steps are required before output is used in client matters — are not just best practices under this framework; they are part of the supervisory structure the rules require.

Fees (NY RPC 1.5)

The opinion addresses the billing question that has generated the most internal debate at firms: can attorneys bill clients for time spent reviewing AI-generated work product? The short answer from the opinion is yes, subject to the reasonableness standard — but attorneys cannot bill for time the AI tool eliminated. Billing a client for four hours of research when an AI tool produced a draft in twenty minutes and the attorney spent thirty minutes reviewing it raises a reasonableness problem under NY RPC 1.5.

The opinion does not resolve every billing scenario, and it does not address whether attorneys must disclose AI use to clients on invoices. That question remains open under New York rules as of this writing.

Candor and Disclosure

The opinion does not impose a blanket disclosure obligation — attorneys are not required to tell clients or courts that they used AI tools simply because they used them. But the candor analysis under NY RPC 3.3 applies in full: if AI-generated content contains errors that reach a tribunal, the attorney's obligation to correct the record is unchanged. And if a client asks whether AI was used, the attorney cannot misrepresent the answer.

Some New York courts have issued standing orders requiring AI disclosure in filings. Those orders operate independently of the NYC Bar opinion and impose their own requirements. Attorneys should check applicable court orders before filing in any New York federal or state court.

How the NYC Bar Opinion Compares to Peer Guidance

What the Opinion Does Not Cover

• Unauthorized practice of law (UPL) issues arising from AI tools marketed directly to non-attorneys — the opinion is scoped to attorney conduct.
• AI use in criminal defense contexts, where additional constitutional considerations may apply.
• Specific product-by-product assessments — the opinion does not evaluate Harvey, Westlaw CoCounsel, Lexis+ AI, or any named tool.
• Court-specific disclosure requirements, which are governed by individual standing orders and local rules rather than bar ethics opinions.
• Whether AI-assisted work product is protected by attorney-client privilege or work product doctrine — those questions are governed by evidence law, not professional conduct rules.

Practical Implications for Attorneys and Firms

The opinion does not prohibit AI use. It structures the conditions under which AI use is ethically permissible, and it assigns responsibility clearly: the attorney is accountable for the output, regardless of which tool produced it.

For solo practitioners and small firms, the most immediate obligation is the confidentiality inquiry. Using a consumer AI tool for client matters without reviewing its data terms is the clearest compliance gap the opinion identifies. Enterprise-tier agreements with legal AI vendors — which typically include zero-retention commitments and data isolation — address this risk, but attorneys should not assume an enterprise agreement exists simply because a vendor markets to law firms.

For larger firms, the supervision framework has the broadest operational impact. Firms that have deployed AI tools without written policies governing review requirements, output verification, and billing treatment are operating without the supervisory structure the opinion implies is required. This is not a theoretical risk — it is the kind of gap that surfaces in malpractice claims and disciplinary proceedings when an AI-generated error reaches a client or court.

Primary Source and Verification

The full text of NYC Bar Formal Opinion 2024-5 is available through the NYC Bar Association's ethics opinions portal. Practitioners should access the primary document directly rather than relying on summaries, including this one. Ethics opinions are occasionally amended or supplemented, and the controlling text is always the version published by the issuing body.