{"id":107916,"date":"2025-02-07T08:24:52","date_gmt":"2025-02-07T16:24:52","guid":{"rendered":"https:\/\/xira.com\/p\/2025\/02\/07\/a-compendium-of-legal-ethics-opinions-on-gen-ai-as-compiled-by-you-guessed-it-gen-ai\/"},"modified":"2025-02-07T08:24:52","modified_gmt":"2025-02-07T16:24:52","slug":"a-compendium-of-legal-ethics-opinions-on-gen-ai-as-compiled-by-you-guessed-it-gen-ai","status":"publish","type":"post","link":"https:\/\/xira.com\/p\/2025\/02\/07\/a-compendium-of-legal-ethics-opinions-on-gen-ai-as-compiled-by-you-guessed-it-gen-ai\/","title":{"rendered":"A Compendium of Legal Ethics Opinions on Gen AI (As Compiled by \u2013 You Guessed It \u2013 Gen AI)"},"content":{"rendered":"

This week, I have been experimenting with Deep Research, the AI agent OpenAI released on Sunday that it says is capable of completing multi-step research tasks and synthesizing large amounts of online information. Not to be confused with the controversial Chinese AI product DeepSeek), Deep Research is said to be particularly useful for people in [\u2026]<\/p>\n

This week, I have been experimenting with Deep Research<\/a>, the AI agent OpenAI released on Sunday that it says is capable of completing multi-step research tasks and synthesizing large amounts of online information. Not to be confused with the controversial Chinese AI product DeepSeek), Deep Research is said to be particularly useful for people in fields such as finance, science and law.<\/p>\n

Already this week, I published two of these experiments. In the first, I used it to analyze the legality of President Trump\u2019s pause of federal grants<\/a>. In about 10 minutes, it produced a 9,000 word detailed memorandum, concluding that the pause \u201cappears to rest on shaky legal ground.\u201d<\/p>\n

Next, I used it to research and recommend the best law practice management suite for a four-lawyer firm. It produced a fairly detailed response<\/a>, including two charts comparing features, pricing, usability, security, support and user satisfaction.<\/p>\n

For today\u2019s task, I asked it to create a report detailing every legal ethics opinion pertaining to generative AI. Here was my exact prompt:<\/p>\n

\u201cCreate a report detailing every legal ethics opinion from every national, state, local and specialty bar association or lawyer licensing body pertaining to the ethics of lawyers\u2019 use of generative artificial intelligence.\u201d<\/p>\n

It responded to my prompt with several questions about the scope of the research I\u2019d requested, such as whether it should focus only on formal ethics opinions or also include informal guidance. After I answered its questions, it produced the report published below. After it produced the report, I asked it to also summarize the findings in a chart, which is what you see immediately below.<\/p>\n

I have not verified that this is a complete list. If anyone knows where I can find a complete list to compare against, please let me know.<\/p>\n

That said, I was again impressed by its ability to conduct comprehensive research across multiple sources and generate a report. The entire task took it 15 minutes.<\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Issuing Body<\/th>\nOpinion Title\/Number<\/th>\nDate Issued<\/th>\nKey Themes<\/th>\nDisclosure Required<\/th>\nBilling Guidance<\/th>\n<\/tr>\n<\/thead>\n
ABA<\/a><\/td>\nFormal Opinion 512<\/td>\nJuly 2024<\/td>\nCompetence, Confidentiality, Supervision, Candor, Fees<\/td>\nCase-dependent<\/td>\nAI efficiency should reduce fees<\/td>\n<\/tr>\n
California<\/a><\/td>\nPractical Guidance<\/td>\nNovember 2023<\/td>\nConfidentiality, Competence, AI Disclosure, Supervision<\/td>\nNot mandatory, but recommended<\/td>\nEfficiency gains must be fair to clients<\/td>\n<\/tr>\n
Florida<\/a><\/td>\nAdvisory Opinion 24-1<\/td>\nJanuary 2024<\/td>\nConfidentiality, Supervision, Fees, AI Transparency<\/td>\nClient informed consent advised<\/td>\nMust not charge for AI time savings<\/td>\n<\/tr>\n
New York State Bar<\/a><\/td>\nTask Force Report<\/td>\nApril 2024<\/td>\nCompetence, Bias, AI Oversight, Client Communication<\/td>\nRecommended in certain cases<\/td>\nClients should benefit from AI-driven efficiency<\/td>\n<\/tr>\n
New York City Bar<\/a><\/td>\nFormal Opinion 2024-5<\/td>\nAugust 2024<\/td>\nConfidentiality, Competence, Supervision, AI Disclosure<\/td>\nRequired for external AI use<\/td>\nNo overcharging for AI use<\/td>\n<\/tr>\n
New Jersey Supreme Court<\/a><\/td>\nPreliminary Guidelines<\/td>\nJanuary 2024<\/td>\nCompetence, Candor, Supervision, AI Verification<\/td>\nNot always, but recommended<\/td>\nNo billing for time not actually spent<\/td>\n<\/tr>\n
Pennsylvania & Philadelphia Bars<\/a><\/td>\nJoint Opinion 2024-200<\/td>\nJune 2024<\/td>\nCompetence, Candor, AI Hallucinations, Billing Ethics<\/td>\nYes, in some AI use cases<\/td>\nCannot bill AI time as manual work<\/td>\n<\/tr>\n
Kentucky<\/a><\/td>\nEthics Opinion KBA E-457<\/td>\nMarch 2024<\/td>\nCompetence, AI Use in Billing, Supervision, Client Consent<\/td>\nDepends on AI\u2019s impact on case<\/td>\nCannot bill for AI learning time<\/td>\n<\/tr>\n
North Carolina<\/a><\/td>\nFormal Ethics Opinion 2024-1<\/td>\nNovember 2024<\/td>\nCompetence, Supervision, Confidentiality, AI Oversight<\/td>\nNot always, but must protect confidentiality<\/td>\nAI-based efficiency should lower costs<\/td>\n<\/tr>\n
Texas<\/a><\/td>\nProposed Opinion 2024-6<\/td>\nNovember 2024<\/td>\nCompetence, Confidentiality, AI Trustworthiness, Supervision<\/td>\nNot explicitly required<\/td>\nFair pricing required for AI use<\/td>\n<\/tr>\n
Virginia<\/a><\/td>\nAI Guidance Update<\/td>\nAugust 2024<\/td>\nConfidentiality, Billing, Supervision, AI Court Compliance<\/td>\nNot mandated but recommended<\/td>\nAI costs must align with ethical billing<\/td>\n<\/tr>\n
D.C. Bar<\/a><\/td>\nEthics Opinion 388<\/td>\nSeptember 2024<\/td>\nCompetence, AI Verification, Supervision, Client Files<\/td>\nRequired in specific situations<\/td>\nNo excess fees for AI use<\/td>\n<\/tr>\n
USPTO<\/a><\/td>\nPractice Guidance (2023\u20132024)<\/td>\nApril 2024<\/td>\nCandor, Confidentiality, AI Use in Legal Submissions<\/td>\nCourt compliance required<\/td>\nLegal AI use cannot inflate costs<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

National Bar Associations<\/strong><\/h2>\n

American Bar Association \u2013 Formal Opinion 512 (July 2024)<\/strong><\/h3>\n

The ABA Standing Committee on Ethics and Professional Responsibility issued Formal Opinion 512, \u201cGenerative Artificial Intelligence Tools,\u201d on July 29, 2024<\/strong>\u200b. americanbar.org<\/a> americanbar.org<\/a>. This is the ABA\u2019s first ethics guidance focused on generative AI use by lawyers. It instructs attorneys to \u201cfully consider their applicable ethical obligations, including their duties to provide competent legal representation, to protect client information, to communicate with clients, to supervise their employees and agents, to advance only meritorious claims and contentions, to ensure candor toward the tribunal, and to charge reasonable fees.\u201d<\/em>\u200b jenkinslaw.org<\/a> In short, existing ABA Model Rules apply to AI just as they do to any technology.<\/p>\n

Key concerns and recommendations:<\/strong> The opinion emphasizes that lawyers must maintain technological competence<\/strong> \u2013 understanding the benefits and risks of AI tools they use\u200b jenkinslaw.org<\/a>. It notes the duty of confidentiality<\/strong> (Model Rule 1.6) requires caution when inputting client data into AI tools; lawyers should ensure no confidential information is revealed without informed client consent\u200b jenkinslaw.org<\/a>. Lawyers should also evaluate whether to inform or obtain consent from clients<\/strong> about AI use, especially if using it in ways that affect the representation\u200b\u00a0jenkinslaw.org<\/a>. AI outputs must be independently verified for accuracy to fulfill duties of candor<\/strong> and avoid filing false or frivolous material (Rules 3.3, 3.1)\u200b jenkinslaw.org<\/a>. The ABA highlights that \u201challucinations\u201d (convincing but false outputs) are a major pitfall\u200b americanbar.org<\/a>. Supervision<\/strong> duties (Rules 5.1 and 5.3) mean lawyers must oversee both subordinate lawyers and nonlawyers and<\/em> the AI tools they use\u200b jenkinslaw.org<\/a>. The opinion also warns that fees<\/strong> must be reasonable \u2013 if AI improves efficiency, lawyers should not overbill for time not actually spent\u200b\u00a0kaiserlaw.com<\/a>. Overall, Formal Op. 512 provides a comprehensive framework mapping generative AI use to existing ethics rules\u200b americanbar.org<\/a> americanbar.org<\/a>.<\/p>\n

(See ABA Formal Op. 512<\/em>\u200b jenkinslaw.org<\/a><\/em> for full text.)<\/em><\/p>\n

State Bar Associations and Regulatory Bodies<\/strong><\/h2>\n

California \u2013 \u201cPractical Guidance\u201d by COPRAC (November 2023)<\/strong><\/h3>\n

The State Bar of California<\/strong> took early action by issuing \u201cPractical Guidance for the Use of Generative AI in the Practice of Law,\u201d<\/strong> approved by the Bar\u2019s Board of Trustees on Nov. 16, 2023\u200b<\/p>\n

calbar.ca.gov<\/a> jdsupra.com<\/a>. Rather than a formal opinion, it is a guidance document (in chart format) developed by the Committee on Professional Responsibility and Conduct (COPRAC). It applies California\u2019s Rules of Professional Conduct to generative AI scenarios.<\/p>\n

Key points:<\/strong> California\u2019s guidance stresses confidentiality<\/strong> \u2013 attorneys \u201cmust not input any confidential client information\u201d<\/em> into AI tools that lack adequate protections\u200b calbar.ca.gov<\/a>. Lawyers should vet an AI vendor\u2019s security and data use policies, and anonymize or refrain from sharing sensitive data<\/strong> unless certain it will be protected\u200b calbar.ca.gov<\/a> calbar.ca.gov<\/a>. The duty of competence and diligence<\/strong> requires understanding how the AI works and its limitations\u200b\u00a0jdsupra.com<\/a>. Lawyers should review AI outputs for accuracy and bias, and \u201cAI should never replace a lawyer\u2019s professional judgment.\u201d<\/em>\u200b jdsupra.com<\/a> If AI assists with research or drafting, the attorney must critically review the results. The guidance also addresses supervision<\/strong>: firms should train and supervise lawyers and staff in proper AI use jdsupra.com<\/a>. Communication with clients<\/strong> may entail disclosing AI use in some cases \u2013 e.g. if it materially affects the representation \u2013 but California did not mandate<\/em> disclosure in all instances\u200b jdsupra.com<\/a>. Finally, the guidance notes candor<\/strong>: the duty of candor to tribunals means attorneys must check AI-generated citations and facts to avoid false statements in court\u200b jdsupra.com<\/a>. Overall, California\u2019s approach is to treat AI as another technology that must be used consistent with existing rules on competence, confidentiality, supervision, etc., providing \u201cguiding principles rather than best practices\u201d\u200b calbar.ca.gov<\/a>.<\/p>\n

(Source: State Bar of CA Generative AI Guidance<\/em>\u200b jdsupra.com<\/a> jdsupra.com<\/a>.)<\/em><\/p>\n

Florida \u2013 Advisory Opinion 24-1 (January 2024)<\/strong><\/h3>\n

The Florida Bar<\/strong> issued Proposed Advisory Opinion 24-1<\/strong> in late 2023, which was adopted by the Bar\u2019s Board of Governors in January 2024\u200b floridabar.org<\/a> floridabar.org<\/a>. Titled \u201cLawyers\u2019 Use of Generative AI,\u201d<\/em> this formal ethics opinion gives a green light to using generative AI \u201cto the extent that the lawyer can reasonably guarantee compliance with the lawyer\u2019s ethical obligations.\u201d\u00a0<\/strong>floridabar.org<\/a> It identifies four focus areas: confidentiality, oversight, fees, and advertising\u200b hinshawlaw.com<\/a> hinshawlaw.com<\/a>.<\/p>\n

Key points:<\/strong> Confidentiality:<\/strong> Florida stresses that protecting client confidentiality (Rule 4-1.6) is paramount. Lawyers should take \u201creasonable steps to prevent inadvertent or unauthorized disclosure\u201d<\/em> of client info by an AI system\u200b jdsupra.com<\/a>. The opinion \u201cadvisable to obtain a client\u2019s informed consent before using a third-party AI that would disclose confidential information.\u201d<\/em>\u200b<\/p>\n

jdsupra.com<\/a> This aligns with prior cloud-computing opinions. Oversight:<\/strong> Generative AI must be treated like a non-lawyer assistant \u2013 the lawyer must supervise and vet its work\u200b jdsupra.com<\/a>. The opinion warns that lawyers relying on AI face \u201cthe same perils as relying on an overconfident nonlawyer assistant\u201d<\/em>\u200b floridabar.org<\/a>. Attorneys must review AI outputs (research, drafts, etc.) for accuracy and legal soundness before use\u200b floridabar.org<\/a>. Notably, after the infamous Mata v. Avianca<\/em> incident of fake cases, Florida emphasizes candor<\/strong>: no frivolous or false material from AI should be submitted\u200b floridabar.org<\/a>. Fees:<\/strong> Improved efficiency from AI cannot be used to charge inflated fees. A lawyer \u201ccan ethically only charge a client for actual costs incurred\u201d<\/em> \u2013 time saved by AI should not be billed as if the lawyer did the work\u200b jdsupra.com<\/a>. If a lawyer will charge for using an AI tool (as a cost), the client must be informed in writing\u200b jdsupra.com<\/a>. And training time<\/strong> \u2013 a lawyer\u2019s time learning an AI tool \u2013 cannot<\/em> be billed to the client\u200b jdsupra.com<\/a>. Advertising:<\/strong> If lawyers advertise their use of AI, they must not be false or misleading. Florida specifically notes that if using a chatbot to interact with potential clients, those users must be told they are interacting with an AI, not a human lawyer\u200b jdsupra.com<\/a>. Any claims about an AI\u2019s capabilities must be objectively verifiable (no puffery that your AI is \u201cbetter\u201d than others without proof)\u200b floridabar.org<\/a> floridabar.org<\/a>. In sum, Florida concludes: \u201ca lawyer may ethically utilize generative AI, but only to the extent the lawyer can reasonably guarantee compliance with duties of confidentiality, candor, avoiding frivolous claims, truthfulness, reasonable fees, and proper advertising.\u201d<\/em>\u200b floridabar.org<\/a>.<\/p>\n

(Sources: Florida Bar Op. 24-1<\/em>\u200b floridabar.org<\/a> jdsupra.com<\/a>.)<\/em><\/p>\n

New York State Bar Association \u2013 Task Force Report (April 2024)<\/strong><\/h3>\n

The New York State Bar Association (NYSBA)<\/strong> did not issue a formal ethics opinion via its ethics committee, but its Task Force on Artificial Intelligence<\/strong> produced a comprehensive 85-page report<\/em> adopted by the House of Delegates on April 6, 2024\u200b floridabar.org<\/a> floridabar.org<\/a>. This report includes a chapter on the \u201cEthical Impact\u201d<\/em> of AI on law practice\u200b floridabar.org<\/a>, effectively providing guidance to NY lawyers. It mirrors many concerns seen in formal opinions elsewhere.<\/p>\n

Key points:<\/strong> The NYSBA report underscores competence<\/strong> and cautions against \u201ctechno-solutionism.\u201d It notes that \u201ca refusal to use technology that makes legal work more accurate and efficient may be considered a refusal to provide competent representation\u201d\u00a0<\/em>nysba.org<\/a> nysba.org<\/a> \u2013 implying lawyers should stay current with helpful AI tools. At the same time, it warns attorneys not to blindly trust AI as a silver bullet. The report coins \u201ctechno-solutionism\u201d<\/em> as the overbelief that new tech (like gen AI) can solve all problems, reminding lawyers that human verification is still required\u200b nysba.org<\/a> nysba.org<\/a>. The infamous Avianca case is cited to illustrate the need to verify AI outputs<\/strong> and supervise the \u201cnonlawyer\u201d tool (AI) under Rule 5.3\u200b nysba.org<\/a>. The report addresses the duty of confidentiality & privacy<\/strong> in depth: Lawyers must ensure client information isn\u2019t inadvertently shared or used to train public AI models\u200b nysba.org<\/a> nysba.org<\/a>. It suggests that if AI tools store or learn from inputs, that raises confidentiality concerns\u200b nysba.org<\/a>. Client consent or use of secure \u201cclosed\u201d AI systems may be needed to protect privileged data. The report also covers supervision<\/strong> (Rule 5.3) \u2013 attorneys should supervise AI use similarly to how they supervise human assistants\u200b nysba.org<\/a>. It touches on bias<\/strong> and fairness, noting generative AI trained on biased data could perpetuate discrimination, which lawyers must guard against\u200b lawnext.com<\/a>. Interestingly, the NYSBA guidance also links AI use to reasonable fees<\/strong>: it suggests effective use of AI can factor into whether a fee is reasonable\u200b jdsupra.com<\/a> jdsupra.com<\/a> (e.g. inefficiently refusing to use available AI might waste client money, whereas using AI and still charging full hours might be unreasonable). In sum, New York\u2019s bar leaders affirm that ethical duties of competence, confidentiality, and supervision fully apply to AI. They encourage using AI\u2019s benefits to improve service, but caution against its risks and urge ongoing attorney oversight floridabar.org<\/a> floridabar.org<\/a>.<\/p>\n

(Sources: NYSBA Task Force Report<\/em>\u200b nysba.org<\/a><\/em> nysba.org<\/a>.)<\/em><\/p>\n

New York City Bar Association \u2013 Formal Opinion 2024-5 (August 2024)<\/strong><\/h3>\n

The New York City Bar Association Committee on Professional Ethics<\/strong> issued Formal Ethics Opinion 2024-5<\/strong> on August 7, 2024\u200b nydailyrecord.com<\/a> nydailyrecord.com<\/a>. This opinion, in a user-friendly chart format, provides practical guidelines for NYC lawyers on generative AI. The Committee explicitly aimed to give \u201cguardrails and not hard-and-fast restrictions\u201d in this evolving area\u200b nydailyrecord.com<\/a>.<\/p>\n

Key points:<\/strong> Confidentiality:<\/strong> The NYC Bar draws a distinction between \u201cclosed\u201d AI systems<\/em> (e.g. an in-house or vendor tool that does not<\/strong> share data externally) and public AI services like ChatGPT. If using an AI that stores or shares inputs outside the firm, client informed consent is required<\/strong> before inputting any confidential information\u200b nydailyrecord.com<\/a>. Even with closed\/internal AI, lawyers must maintain internal confidentiality protections. The opinion warns lawyers to review AI Terms of Use<\/strong> regularly to ensure the provider isn\u2019t using or exposing client data without consent\u200b\u00a0nydailyrecord.com<\/a>. Competence:<\/strong> Echoing others, NYC advises that lawyers \u201cunderstand to a reasonable degree how the technology works, its limitations, and the applicable Terms of Use\u201d<\/em> before using generative AI\u200b nydailyrecord.com<\/a>. Attorneys should avoid delegating their professional judgment to AI; any AI output is just a starting point or draft\u200b nydailyrecord.com<\/a>. Lawyers must ensure outputs are accurate and tailored to the client\u2019s needs<\/strong> \u2013 essentially, verify everything<\/em> and edit AI-generated material so that it truly serves the client\u2019s interests\u200b nydailyrecord.com<\/a>. Supervision:<\/strong> Firms should implement policies and training for lawyers and staff on acceptable AI use\u200b nydailyrecord.com<\/a>. The Committee notes that client intake chatbots<\/strong> (if used on a firm\u2019s website, for example) require special oversight to avoid inadvertently forming attorney-client relationships or giving legal advice without proper vetting\u200b nydailyrecord.com<\/a>. In other words, a chatbot interacting with the public should be carefully monitored by lawyers to ensure it doesn\u2019t mislead users about its nature or create unintended obligations\u200b nydailyrecord.com<\/a>. The NYC Bar\u2019s guidance aligns with California\u2019s in format and substance, reinforcing that the core duties of confidentiality, competence (tech proficiency), and supervision all apply when lawyers use generative AI tools\u200b nydailyrecord.com<\/a> nydailyrecord.com<\/a>.<\/p>\n

(Source: NYC Bar Formal Op. 2024-5<\/em>\u200bnydailyrecord.com<\/a><\/em> nydailyrecord.com<\/a>.)<\/em><\/p>\n

New Jersey Supreme Court \u2013 Preliminary Guidelines (January 2024)<\/strong><\/h3>\n

In New Jersey, the state\u2019s highest court itself weighed in. On January 24, 2024, the New Jersey Supreme Court\u2019s Committee on AI and the Courts<\/strong> issued \u201cPreliminary Guidelines on the Use of AI by New Jersey Lawyers,\u201d<\/strong> which were published as a Notice to the Bar\u200b njcourts.gov<\/a> njcourts.gov<\/a>. These guidelines, effective immediately, aim to help NJ lawyers comply with existing Rules of Professional Conduct when using generative AI\u200b njcourts.gov<\/a>.\u00a0<\/strong><\/p>\n

Key points:<\/strong> The Court made clear that AI does not change lawyers\u2019 fundamental duties<\/strong>. Any use of AI \u201cmust be employed with the same commitment to diligence, confidentiality, honesty, and client advocacy as traditional methods of practice.\u201d<\/em>\u200b njcourts.gov<\/a> In other words, tech advances do not<\/em> dilute responsibilities. The NJ guidelines highlight accuracy and truthfulness<\/strong>: lawyers have an ethical duty to ensure their work is accurate, so they must always check AI-generated content for \u201challucinations\u201d or errors<\/em> before relying on it\u200b jdsupra.com<\/a>. Submitting false or fake information generated by AI would violate rules against misrepresentations to the court. The guidelines reiterate candor to tribunals<\/strong> \u2013 attorneys must not present AI-produced output containing fabricated cases or facts (the Mata\/Avianca<\/em> situation is alluded to)\u200bjdsupra.com<\/a>. Regarding communication and client consent<\/strong>, NJ took a measured approach: There is \u201cno per se requirement to inform a client\u201d<\/em> about every AI use, unless<\/em> not telling the client would prevent the client from making informed decisions about the representation\u200b jdsupra.com<\/a>. For example, if AI is used in a trivial manner (typo correction, formatting), disclosure isn\u2019t required; but if it\u2019s used in substantive tasks that affect the case, lawyers should consider informing the client, especially if there\u2019s heightened risk. Confidentiality:<\/strong> Lawyers must ensure any AI tool is secure to avoid inadvertent disclosures of client info\u200b jdsupra.com<\/a>. This echoes the duty to use \u201creasonable efforts\u201d<\/em> to safeguard confidential data (RPC 1.6). No misconduct:<\/strong> The Court reminds that all rules on attorney misconduct (dishonesty, fraud, bias, etc.) apply in AI usage\u200b jdsupra.com<\/a>. For instance, using AI in a way that produces discriminatory outcomes or that frustrates justice would breach Rule 8.4. Supervision:<\/strong> Law firms must supervise how their lawyers and staff use AI jdsupra.com<\/a> \u2013 establishing internal policies to ensure ethical use. Overall, New Jersey\u2019s top court signaled that it embraces innovation (noting AI\u2019s potential benefits) but insists lawyers \u201cbalance the benefits of innovation while safeguarding against misuse.\u201d<\/em>\u200b njcourts.gov<\/a><\/p>\n

(Sources: NJ Supreme Court Guidelines<\/em>\u200b jdsupra.com<\/a> jdsupra.com<\/a>.)<\/em><\/p>\n

Pennsylvania & Philadelphia Bars \u2013 Joint Opinion 2024-200 (June 2024)<\/strong><\/h3>\n

The Pennsylvania Bar Association (PBA)<\/strong> and Philadelphia Bar Association<\/strong> jointly issued Formal Opinion 2024-200<\/strong> in mid-2024\u200b lawnext.com<\/a> lawnext.com<\/a>. This collaborative opinion (\u201cJoint Formal Op. 2024-200\u201d) provides ethical guidance for Pennsylvania lawyers using generative AI. It repeatedly emphasizes that the same rules apply to AI as to any technology<\/em>\u200b lawnext.com<\/a>.<\/p>\n

Key points:<\/strong> The joint opinion places heavy emphasis on competence (Rule 1.1)<\/strong>. It famously states \u201cLawyers must be proficient in using technological tools to the same extent they are in traditional methods\u201d<\/em>\u200b lawnext.com<\/a>. In other words, attorneys should treat AI as part of the competence duty \u2013 understanding e-discovery software, legal research databases, and now generative AI<\/em>, is part of being a competent lawyer\u200b lawnext.com<\/a>. The opinion acknowledges generative AI\u2019s unique risk: it can hallucinate<\/strong> (generate false citations or facts)\u200b lawnext.com<\/a>. Thus, due diligence is required \u2013 lawyers must verify all AI outputs<\/strong>, especially legal research results and citations\u200b lawnext.com<\/a>\u00a0lawnext.com<\/a>. The opinion bluntly warns that if you ask AI for cases and \u201cthen file them in court without even bothering to read or Shepardize them, that is stupid.\u201d<\/em>\u200b lawnext.com<\/a> (The opinion uses more polite language, but this captures the spirit.) It highlights bias<\/strong> as well: AI may carry implicit biases from training data, so lawyers should be alert to any discriminatory or skewed content in AI output\u200b lawnext.com<\/a>. The Pennsylvania\/Philly opinion also advises lawyers to communicate with clients<\/strong> about AI use. Specifically, lawyers should be transparent and \u201cprovide clear, transparent explanations\u201d<\/em> of how AI is being used in the case\u200b lawnext.com<\/a> lawnext.com<\/a>. In some situations, obtaining client consent<\/em> before using certain AI tools is recommended\u200b lawnext.com<\/a> lawnext.com<\/a> \u2013 e.g., if the tool will handle confidential information or significantly shape the legal work. The opinion lays out \u201c12 Points of Responsibility\u201d<\/strong> for using gen AI\u200b lawnext.com<\/a> lawnext.com<\/a>, which include many of the above: ensure truthfulness and accuracy of AI-derived content, double-check citations, maintain confidentiality (ensure AI vendors keep data secure)\u200b lawnext.com<\/a>, check for conflicts (make sure use of AI doesn\u2019t introduce any conflict of interest)\u200b lawnext.com<\/a>, and transparency<\/strong> with clients, courts, and colleagues about AI use and its limitations\u200b lawnext.com<\/a>. It also addresses proper billing practices<\/strong>: lawyers shouldn\u2019t overcharge when AI boosts efficiency\u200b lawnext.com<\/a>. If AI saves time, the lawyer should not bill as if they did it manually \u2013 they may bill for the actual<\/em> time or consider value-based fees, but padding hours violates the rule on reasonable fees\u200b lawnext.com<\/a>. Overall, the Pennsylvania and Philly bars take the stance that embracing AI is fine \u2014 even beneficial \u2014 as long as lawyers \u201cremain fully accountable for the results,\u201d<\/em> use AI carefully, and don\u2019t neglect any ethical duty in the process\u200b lawnext.com<\/a> lawnext.com<\/a>.<\/p>\n

(Sources: Joint PBA\/Phila. Opinion 2024-200 summarized by Ambrogi<\/em>\u200b lawnext.com<\/a><\/em> lawnext.com<\/a>.)<\/em><\/p>\n

Kentucky \u2013 Ethics Opinion KBA E-457 (March 2024)<\/strong><\/h3>\n

The Kentucky Bar Association<\/strong> issued Ethics Opinion KBA E-457, \u201cThe Ethical Use of Artificial Intelligence in the Practice of Law,\u201d<\/strong> on March 15, 2024\u200b cdn.ymaws.com<\/a>. This formal opinion (finalized after a comment period in mid-2024) provides a nuanced roadmap for Kentucky lawyers. It not only answers basic questions but also offers broader insight, reflecting the work of a KBA Task Force on AI\u200b techlawcrossroads.com<\/a>.<\/p>\n

Key points:<\/strong> Competence:<\/strong> Like other jurisdictions, Kentucky affirms that keeping abreast of technology (including AI) is a mandatory<\/em> aspect of competence\u200b techlawcrossroads.com<\/a> techlawcrossroads.com<\/a>. Kentucky\u2019s Rule 1.1 Comment 6 (equivalent to ABA Comment 8) says lawyers \u201cshould keep abreast of \u2026 the benefits and risks associated with relevant technology.\u201d<\/em> The opinion stresses this is not optional: \u201cIt\u2019s not a \u2018should\u2019; it\u2019s a must.\u201d<\/em>\u200b techlawcrossroads.com<\/a> Lawyers cannot ethically ignore AI\u2019s existence or potential in law practice\u200b techlawcrossroads.com<\/a> techlawcrossroads.com<\/a> (implying that failing to understand how AI might improve service could itself be a lapse in competence). Disclosure to clients:<\/strong> Kentucky takes a practical stance that there is \u201cno duty to disclose to the client the \u2018rote\u2019 use of AI generated research,\u201d<\/em> absent special circumstances\u200b techlawcrossroads.com<\/a>. If an attorney is just using AI as a tool (like one might use Westlaw or a spell-checker), they generally need not inform the client. However, there are important exceptions<\/em> \u2013 if the client has specifically limited use of AI, or if use of AI presents significant risk or would require client consent under the rules, then disclosure is needed techlawcrossroads.com<\/a>. Lawyers should discuss risks and benefits<\/strong> of AI with clients if client consent is required for its use (for example, if AI will process confidential data, informed consent may be wise)\u200b techlawcrossroads.com<\/a>. Fees:<\/strong> KBA E-457 is very direct about fees and AI. If AI significantly reduces the time spent on a matter, the lawyer may need to reduce their fees<\/em> accordingly\u200b techlawcrossroads.com<\/a>. A lawyer cannot charge a client as if a task took 5 hours if AI allowed it to be done in 1 hour \u2013 that would make the fee unreasonable. The opinion also says a lawyer can only charge a client for the expense<\/em> of using AI (e.g., the cost of a paid AI service) if the client agrees to that fee in writing\u200b techlawcrossroads.com<\/a>. Otherwise, passing along AI tool costs may be impermissible. In short, AI\u2019s efficiencies should benefit clients, not become a hidden profit center. Confidentiality:<\/strong> Lawyers have a \u201ccontinuing duty to safeguard client information if they use AI,\u201d<\/em> and must comply with all applicable court rules on AI use\u200b techlawcrossroads.com<\/a>. This means vetting AI providers\u2019 security and ensuring no confidential data is exposed. Kentucky echoes that attorneys must understand the terms and operation<\/strong> of any third-party AI system they use\u200b techlawcrossroads.com<\/a>. They should know how the AI service stores and uses data. Court rules compliance:<\/strong> Notably, the opinion reminds lawyers to follow any court-imposed rules about AI (for instance, if a court requires disclosure of AI-drafted filings, the lawyer must do so)\u200b cdn.ymaws.com<\/a>. Firm policies and training:<\/strong> KBA E-457 advises law firms to create informed policies on AI use and to supervise those they manage in following these policies\u200b techlawcrossroads.com<\/a>. In summary, Kentucky\u2019s opinion encourages lawyers to embrace AI\u2019s potential but to do so carefully: stay competent with the technology, be transparent when needed, adjust fees fairly, protect confidentiality, and always maintain ultimate responsibility for the work. It concludes that Kentucky lawyers \u201ccannot run from or ignore AI.\u201d<\/em>\u200b techlawcrossroads.com<\/a><\/p>\n

(Source: KBA E-457 (2024) via TechLaw Crossroads summary<\/em> techlawcrossroads.com<\/a><\/em> techlawcrossroads.com<\/a>.)<\/em><\/p>\n

North Carolina \u2013 Formal Ethics Opinion 2024-1 (November 2024)<\/strong><\/h3>\n

The North Carolina State Bar<\/strong> adopted 2024 Formal Ethics Opinion 1, \u201cUse of Artificial Intelligence in a Law Practice,\u201d<\/strong> on November 1, 2024\u200b ncbar.gov<\/a> ncbar.gov<\/a>. This opinion squarely addresses whether and how NC lawyers can use AI tools consistent with their ethical duties.<\/p>\n

Key points:<\/strong> The NC State Bar gives a cautious \u201cYes\u201d<\/strong> to using AI, under specific conditions: \u201cYes, provided the lawyer uses any AI program, tool, or resource competently, securely to protect client confidentiality, and with proper supervision when relying on the AI\u2019s work product.\u201d<\/em>\u200b ncbar.gov<\/a>. That single sentence captures the three pillars of NC\u2019s guidance: competence<\/strong>, confidentiality<\/strong>, and supervision<\/strong>. NC acknowledges that nothing in the Rules explicitly prohibits AI use<\/em>\u200b ncbar.gov<\/a>, so it comes down to applying existing rules. Competence:<\/strong> Lawyers must understand the technology sufficiently to use it effectively and safely\u200b ncbar.gov<\/a>. Rule 1.1 and its Comment in NC (which, like the ABA, includes tech competence) require lawyers to know what they don\u2019t know \u2013 if a lawyer isn\u2019t competent with an AI tool, they must get up to speed or refrain. NC emphasizes that using AI is often the lawyer\u2019s own decision<\/em> but it must be made prudently, considering factors like the tool\u2019s reliability and cost-benefit for the client\u200b ncbar.gov<\/a> ncbar.gov<\/a>. Confidentiality & Security:<\/strong> Rule 1.6(c) in North Carolina obligates lawyers to take reasonable efforts to prevent unauthorized disclosure of client info. So, before using any cloud-based or third-party AI, the lawyer must ensure it is \u201csufficiently secure and compatible with the lawyer\u2019s confidentiality obligations.\u201d<\/em>\u200b ncbar.gov<\/a> ncbar.gov<\/a>. The opinion suggests attorneys evaluate providers like they would any vendor handling client data \u2013 e.g., examine terms of service, data storage policies, etc., similar to prior NC guidance on cloud computing\u200b ncbar.gov<\/a> ncbar.gov<\/a>. If the AI is \u201cself-learning\u201d<\/em> (using inputs to improve itself), lawyers should be wary that client data might later resurface to others\u200b ncbar.gov<\/a>. NC stops short of mandating client consent for AI use, but it implies that if an AI tool can\u2019t be used consistent with confidentiality, then either don\u2019t use it or get client permission. Supervision and Independent Judgment:<\/strong> NC treats AI output like work by a nonlawyer assistant. Under Rule 5.3, lawyers must supervise the use of AI tools and \u201cexercise independent professional judgment in determining how (or if) to use the product of an AI tool\u201d<\/em> for a client\u200b ncbar.gov<\/a> ncbar.gov<\/a>. This means a lawyer cannot blindly accept an AI\u2019s result \u2013 they must review and verify it before relying on it. If an AI drafts a contract or brief, the lawyer is responsible for editing and ensuring it\u2019s correct and appropriate. NC explicitly analogizes AI to both other software and to nonlawyer staff: AI is \u201cbetween\u201d a software tool and a nonlawyer assistant in how we think of it\u200b ncbar.gov<\/a>. Thus, the lawyer must both know how to use the software<\/strong> and supervise its output<\/strong> as if it were a junior employee\u2019s work. Bottom line:<\/strong> NC FO 2024-1 concludes that a lawyer may<\/em> use AI in practice \u2013 for tasks like document review, legal research, drafting, etc. \u2013 as long as the lawyer remains fully responsible for the outcome\u200b ncbar.gov<\/a> ncbar.gov<\/a>. The opinion purposefully doesn\u2019t dictate when AI is appropriate or not, recognizing the technology is evolving\u200b ncbar.gov<\/a>. But it clearly states that if a lawyer decides<\/strong> to employ AI, they are \u201cfully responsible\u201d for its use and must ensure it is competent use, confidential use, and supervised use\u200b ncbar.gov<\/a> ncbar.gov<\/a>.<\/p>\n

(Source: NC 2024 FEO-1<\/em>\u200bncbar.gov<\/a><\/em> ncbar.gov<\/a>.)<\/em><\/p>\n

Texas \u2013 Proposed Opinion 2024-6 (Draft, November 2024)<\/strong><\/h3>\n

The State Bar of Texas Professional Ethics Committee<\/strong> has circulated a Proposed Ethics Opinion No. 2024-6<\/strong> (posted for public comment on Nov. 19, 2024) regarding lawyers\u2019 use of generative AI\u200b\u00a0texasbar.com<\/a>. (As of this writing, it is a draft opinion awaiting final adoption.)<\/em> This Texas draft provides a \u201chigh-level overview\u201d of ethical issues raised by AI, requested by a Bar task force on AI\u200b\u00a0texasbar.com<\/a>.<\/p>\n

Key points (draft):<\/strong> The proposed Texas opinion covers familiar ground. It notes the duty of competence (Rule 1.01)<\/strong> extends to understanding relevant technology\u200b texasbar.com<\/a>. Texas specifically cites its prior ethics opinions on cloud computing and metadata, which required lawyers to have a \u201creasonable and current understanding\u201d<\/em> of those technologies\u200b texasbar.com<\/a> texasbar.com<\/a>. By analogy, any Texas lawyer using generative AI \u201cmust have a reasonable and current understanding of the technology\u201d<\/em> and its capabilities and limits\u200b texasbar.com<\/a>. In practical terms, this means lawyers should educate themselves on how tools like ChatGPT actually work (e.g. that they predict text rather than retrieve vetted sources) and what their known pitfalls are texasbar.com<\/a>. The draft opinion spends time describing Mata v. Avianca<\/em> to illustrate the dangers of not understanding AI\u2019s lack of a reliable legal database\u200b texasbar.com<\/a> texasbar.com<\/a>. On confidentiality (Rule 1.05 in Texas)<\/strong>, the opinion again builds on prior guidance: lawyers must safeguard client information when using any third-party service\u200b texasbar.com<\/a> texasbar.com<\/a>. It suggests precautions similar to those for cloud storage: \u201cacquire a general understanding of how the technology works; review (and potentially renegotiate) the Terms of Service; [ensure] the provider will keep data confidential; and stay vigilant about data security.\u201d<\/em>\u200b texasbar.com<\/a>. (These examples are drawn from Texas Ethics Op. 680 on cloud computing, which the AI opinion heavily references.) If an AI tool cannot be used in a way that protects confidential info, the lawyer should not use it for those purposes. The Texas draft also flags duty to avoid frivolous submissions (Rule 3.01)<\/strong> and duty of candor to tribunal (Rule 3.03)<\/strong> as directly relevant\u200b texasbar.com<\/a>. Using AI doesn\u2019t excuse a lawyer from these obligations \u2013 citing fake cases or making false statements is no less an ethical violation because an AI generated them. Lawyers must thoroughly vet AI-generated legal research and content to ensure it\u2019s grounded in real law and facts\u200b texasbar.com<\/a> texasbar.com<\/a>. The opinion essentially says: if you choose to use AI, you must double-check its work just as you would a junior lawyer\u2019s memo or a nonlawyer assistant\u2019s draft<\/em>. Supervision (Rules 5.01, 5.03):<\/strong> Supervising partners should have firm-wide measures so that any use of AI by their team is ethical texasbar.com<\/a> texasbar.com<\/a>. This could mean creating policies on approved AI tools and requiring verification of AI outputs. In summary, the Texas proposed opinion doesn\u2019t ban generative AI; it provides a \u201csnapshot\u201d of issues and reinforces that core duties of competence, confidentiality, candor, and supervision must guide any use of AI in practice texasbar.com<\/a> texasbar.com<\/a>. (The committee acknowledges the AI landscape is rapidly changing, so they focused on broad principles rather than specifics that might soon be outdated\u200b texasbar.com<\/a>.) Once finalized, Texas\u2019s opinion will likely align with the consensus: lawyers can harness AI\u2019s benefits if they remain careful and accountable.<\/p>\n

(Source: Texas Proposed Op. 2024-6<\/em>\u200b texasbar.com<\/a><\/em> texasbar.com<\/a>.)<\/em><\/p>\n

Virginia State Bar \u2013 AI Guidance Update (August 2024)<\/strong><\/h3>\n

In 2024 the Virginia State Bar<\/strong> released a short set of guidelines on generative AI<\/strong> as an update on its website (around August 2024)\u200b nydailyrecord.com<\/a>. This concise guidance stands out for its practicality and flexibility. Rather than an extensive opinion, Virginia issued overarching advice that can adapt as AI technology evolves\u200b nydailyrecord.com<\/a>.\u00a0<\/strong><\/p>\n

Key points:<\/strong> Virginia first emphasizes that lawyers\u2019 basic ethical responsibilities \u201chave not changed\u201d due to AI<\/strong>, and that generative AI presents issues \u201cfundamentally similar\u201d to those with other technology or with supervising people\u200b nydailyrecord.com<\/a>. This frames the guidance: existing rules suffice. On confidentiality<\/strong>, the Bar advises lawyers to vet how AI providers handle data just as they would with any vendor\u200b nydailyrecord.com<\/a> nydailyrecord.com<\/a> . Legal-specific AI products (designed for lawyers, with better data security) may offer more protection, but even then attorneys \u201cmust make reasonable efforts to assess\u201d<\/em> the security and \u201cwhether and under what circumstances\u201d<\/em> confidential info could be exposed\u200b nydailyrecord.com<\/a>. In other words, even if using an AI tool marketed as secure for lawyers, you should confirm that it truly keeps your client\u2019s data confidential (no sharing or training on it without consent)\u200b nydailyrecord.com<\/a> nydailyrecord.com<\/a>. Virginia notably aligns with most jurisdictions (and diverges from a stricter ABA stance) regarding client consent<\/strong>: \u201cthere is no per se requirement to inform a client about the use of generative AI in their matter\u201d<\/em>\u200b nydailyrecord.com<\/a>. Unless something about the AI use would necessitate client disclosure (e.g., an agreement with the client, or an unusual risk like using a very public AI for sensitive info), lawyers generally need not obtain consent for routine AI use\u200b nydailyrecord.com<\/a>. This is consistent with the idea that using AI can be like using any software tool behind the scenes. Next, supervision and verification<\/strong>: The bar stresses that lawyers must review all AI outputs as they would work done by a junior attorney or nonlawyer assistant\u200b nydailyrecord.com<\/a> nydailyrecord.com<\/a>. Specifically, \u201cverify that any citations are accurate (and real)\u201d<\/em> and generally ensure the AI\u2019s work product is correct\u200b nydailyrecord.com<\/a>. This duty extends to supervising others in the firm \u2013 if a paralegal or associate uses AI, the responsible lawyer must ensure they are doing so properly\u200b\u00a0nydailyrecord.com<\/a>. On fees and billing<\/strong>, Virginia takes a clear stance: a lawyer may not bill a client for time not actually spent<\/em> due to AI efficiency gains\u200b nydailyrecord.com<\/a>. \u201cA lawyer may not charge an hourly fee in excess of the time actually spent \u2026 and may not bill for time saved by using generative AI.\u201d<\/em>\u200b nydailyrecord.com<\/a> If AI cuts a research task from 5 hours to 1, you can\u2019t still charge 5 hours. The Bar suggests considering alternative fee arrangements to account for AI\u2019s value, instead of hourly billing windfalls\u200b nydailyrecord.com<\/a>. As for passing along AI tool costs: the Bar says you can\u2019t charge the client for your AI subscription or usage unless<\/em> it\u2019s a reasonable charge and permitted by the fee agreement\u200b nydailyrecord.com<\/a>. Finally, Virginia reminds lawyers to stay aware of any court rules about AI<\/strong>. Some courts (even outside Virginia) have begun requiring attorneys to certify that filings were checked for AI-generated falsehoods, or even prohibiting AI-drafted documents absent verification. Virginia\u2019s guidance highlights that lawyers must comply with any such disclosure or anti-AI rules in whatever jurisdiction they are in\u200b nydailyrecord.com<\/a> nydailyrecord.com<\/a>. Overall, the Virginia State Bar\u2019s message is: use common sense and existing rules. Be transparent when needed, protect confidentiality, supervise and double-check AI outputs, bill fairly, and follow any new court requirements\u200b nydailyrecord.com<\/a> nydailyrecord.com<\/a>. This short-form guidance was praised for being \u201cstreamlined\u201d and adaptable as AI tools continue to change\u200b nydailyrecord.com<\/a>.<\/p>\n

(Source: Virginia State Bar AI Guidance via N.Y. Daily Record<\/em>\u200b nydailyrecord.com<\/a><\/em> nydailyrecord.com<\/a>.)<\/em><\/p>\n

District of Columbia Bar \u2013 Ethics Opinion 388 (September 2024)<\/strong><\/h3>\n

The D.C. Bar<\/strong> issued Ethics Opinion 388: \u201cAttorneys\u2019 Use of Generative AI in Client Matters\u201d<\/strong> in 2024 (the second half of the year)\u200b kaiserlaw.com<\/a>. This opinion closely analyzes the ethical implications of lawyers using gen AI, using the well-known Mata v. Avianca<\/em> incident as a teaching example\u200b kaiserlaw.com<\/a> kaiserlaw.com<\/a> . It then organizes guidance under specific D.C. Rules of Professional Conduct.\u00a0<\/strong><\/p>\n

Key points:<\/strong> The opinion breaks its analysis into categories of duties\u200b kaiserlaw.com<\/a> kaiserlaw.com<\/a>:<\/p>\n