April 2024 Ethically Speaking - The California State Bar’s New Practical Ethical Guidance on Generative Artificial Intelligence

by Carole J. Buckner

The advent of generative artificial intelligence (GenAI) promises to transform the way we practice law. The California State Bar’s Committee on Professional Responsibility and Conduct (COPRAC) recently published the first guidance for lawyers in California on the use of GenAI. Practical Guidance For the Use of GenAI in the Practice of Law, available at https://www.calbar.ca.gov/Portals/0/documents/ethics/Generative-AI-Practical-Guidance.pdf (the “Guidance”). This article, which provides an overview of the Guidance, was written by an actual human, not using GenAI. We also examine Ethics Opinion 24-1 from the Florida Bar Association (the “FBA Opinion”) (available at https://www.google.com/search?q=florida+bar+ethics+opininon+on+generative+AI&rlz=1C1GCGA_enUS1091US1091&oq=florida+bar+ethics+opininon+on+generative+AI&gs_lcrp=EgZjaHJvbWUyBggAEEUYOTIJCAEQIRgKGKABMgkIAhAhGAoYoAEyCQgDECEYChigAdIBCTE2MDAwajBqN6gCALACAA&sourceid=chrome&ie=UTF-8) which provides additional insight on several topics.

The use of GenAI implicates multiple ethical obligations including the duties of competence, confidentiality, compliance with the law, communication, and candor to the court. Ethical prohibitions on discrimination and harassment by lawyers can be triggered by the use of GenAI. How a lawyer charges clients for the use of GenAI also depends upon a lawyer’s ethical obligations. Lawyers also have duties to supervise the use of GenAI and establish policies governing the use of GenAI. More so than many historic technologies that have been implemented by legal professionals, GenAI itself, and hence the California State Bar’s Guidance, proves to be quite a bit to get one’s head around.

Standard of Care
The challenges presented by GenAI do not mean a lawyer should avoid the use of GenAI entirely, as some lawyers historically may have tried to do with other newfangled technology. With a myriad of GenAI offerings promising to transform the practice of law, many lawyers are already using GenAI, and the failure to do so will, at some unknown future time, put a lawyer at risk of falling below the standard of care.

The failure to implement GenAI might, in the future, result in overbilling, to the extent that legal work is done “manually” rather than by GenAI, on the theory that lawyers should not continue billing on an hourly basis for manual work that could be done reliably using GenAI. Superior functionality could also be available for certain applications. For example, should we be asking GenAI to rewrite our arguments to make them more concise, more persuasive, or more targeted toward the thinking of the specific judge we are in front of on a given case? Presently, we are still at the stage where there is a “lack of clarity as to how [GenAI] works.” Guidance at 1. We have witnessed examples of GenAI’s hallucination and unreliability, with resulting errors creating a risk of discipline, sanctions, and adverse publicity.

Despite this, we must forge ahead into this challenging technological thicket and endeavor to develop an understanding of GenAI while bearing in mind our traditional, well-known ethical duties.

Policy and Supervision
The Guidance is organized according to the applicable ethics rules and statutory requirements set forth in the State Bar Act. A foundational compliance burden falls upon lawyers serving as supervisors and managers who, consistent with Rules 5.1, 5.2, and 5.3 of the Rules of Professional Conduct (references are to the California Rules of Professional Conduct throughout), are obligated as managers and supervisory lawyers to establish “clear policies” regarding permissible use of GenAI, and take measures to assure that such policies comport with lawyers’ ethical obligations, to include training on the use, practical aspects, pitfalls, and ethical issues involved. Faced with the daunting requirement that a manager or supervisor of lawyers must develop a “policy,” one approach that may cross a lawyer’s mind would be to simply ban the use of GenAI in the practice of law. But a more balanced approach would allow GenAI to be used within specified parameters. Whatever the policy is, according to the State Bar, it must be “clear.”

One issue is whether the obligations to supervise non-lawyers include supervising GenAI itself. Rule 5.3 speaks to supervision of “nonlawyers,” and the comment provides examples of humans. The Guidance does not specifically speak to whether there is a duty of supervision over GenAI per se, such as a ChatBot used for intake that leverages GenAI. However, such an obligation may fall within a lawyer’s duty of competence, as discussed below. The FBA Opinion raises the issue of whether AI is a person, in terms of the duty of supervision arising under Rule 5.3, indicating that website chatbots should clearly identify their non-lawyer status to prospective clients, not offer legal advice to prospective clients, and be wary of utilizing an overly welcoming GenAI chatbot that may provide legal advice, fail to identify itself as a chatbot, and provide appropriate disclaimers. FBA Opinion at 1. Although the Guidance does not touch upon chatbots, California law provides that it is unlawful for a person to use a bot to communicate or interact with a person online with intent to mislead the other person about its artificial identity. Cal. Bus. & Prof. Code § 17941.

A core concern arising from the use of GenAI is the lawyer’s duty of confidentiality. See Cal. Bus. & Prof. Code § 6068(e); see also Rule 1.6. This duty prohibits disclosure of confidential information relating to the representation of a client without informed consent of the client. To meet this obligation, the Guidance provides two paths: first, you may decide not to share confidential information with GenAI; alternatively, you may share confidential information in connection with the use of GenAI if the information is not further shared or used for learning, and the system is secure.

As for cybersecurity, can you accept the representations of the vendor? The State Bar’s Guidance indicates that a lawyer should “consult with IT professionals or cybersecurity experts to ensure that any AI system in which a lawyer would input confidential information adheres to stringent security, confidentiality and data retention protocols.” Guidance at 2. Reliance on the GenAI vendor’s statements regarding security, or even a third party’s ostensible verification, is insufficient. With a wide variety of new GenAI applications, this complex and expensive standard may lead a lawyer to decide to take the other path, and decide that it is easier not to share any confidential information. Some GenAI applications will undoubtedly meet this standard.

Another aspect of confidentiality involves how information input into the GenAI program is used within the program, including whether the information is used to train the GenAI program itself, and whether information input into the system is shared later with third parties. This is in part a matter of understanding exactly how the GenAI you are using functions, discussed further below. If the GenAI lacks adequate protocols, or you find it impossible to verify, due to the GenAI’s lack of transparent disclosures, and you decide to take the other path, you “must anonymize client information and avoid entering details that can be used to identify the client.” Guidance at 2.

Competence & Diligence
A foundational element of any GenAI policy should be based on the concept that the lawyer understands the technology and how it works, including its benefits and risks, including use of client information, confidentiality, and security. These obligations arise from the duty of competence set forth in Rule 1.1. Competence implies knowledge, skill, thoroughness, and preparation. The duty of diligence also governs the use of GenAI. Guidance at 2-3.

A starting point is to understand the terms and conditions governing the use of the technology. Guidance at 2. What use is made of the information input into the GenAI software? How is the system protected and what security is employed? In a practical sense, this understanding of risk includes the risk of false information being generated, as has been highly publicized, as well as the risk of biased information being generated. One GenAI program prompted to visually depict a meeting of lawyers reportedly created an image composed of all white males. The Guidance points out compliance with Rule 8.4.1 is required, which prohibits lawyers from engaging in discrimination, and the potential for bias to materialize when using GenAI, for example, when GenAI is used in client screening or employee screening. Guidance at 4.

The Guidance provides that each individual lawyer is responsible for the use of GenAI. Guidance at 3. We are learning that certain types of GenAI are not entirely trustworthy even though the results may project a false “confidence.” The Guidance indicates that we cannot delegate our professional judgment to technology and that we need to avoid “overreliance” on GenAI, which can mislead the court and result in harm to clients. The Guidelines characterize GenAI as a “starting point” to be “critically analyzed for accuracy” by a lawyer who must “critically review, validate and correct” the input and output to assure accurate content that best serves the client’s interests. The Guidance indicates that lawyers “may” supplement GenAI research and argument with “human-performed” research, analysis, and review of legal authorities. As we know from several well-publicized examples, the review of any case citations must be double-checked, as some GenAI programs may generate artificial cases that can appear authentic.

Compliance with Law
Another foundational principle is that the use of GenAI must comply with the law. Guidance at 3. This obligation is triggered by the prohibition on lawyers counseling clients to engage in or assist a client with conduct that the lawyer knows is a violation of the law. Cal. Bus. & Prof. Code § 6068(a); Rules 8.4 & 1.2.1. The use of GenAI may involve a complex analysis of a wide range of applicable laws. The Guidance references some of the multiple areas of the law that may be implicated by the use of GenAI, including artificial intelligence law (for example the EU recently agreed to adopt the Artificial Intelligence Act), as well as existing privacy law, intellectual property, and cybersecurity. Lawyers advising clients regarding the use of GenAI must stay abreast of this rapidly changing area and must comply with the applicable regulations regarding the use of GenAI as it is implemented in their practices.

A lawyer’s duty of candor to the court applies to the use of GenAI. Guidance at 4. Lawyers must avoid misleading the court about the facts or the law, as well as offering false evidence. Rule 3.3. Because GenAI may generate inaccurate facts or law, everything must be checked by the lawyer. We know of some hallucinations in which GenAI created fictional legal citations for cases that did not actually exist. Should this occur, a lawyer is required to correct any erroneous information a lawyer presents to the court.

GenAI might also create facts that are not accurate. One GenAI program prompted to write a marketing bio for a lawyer touted the fact that the person being profiled was an accomplished marathon runner, although that was not the case. The Guidance also cautions that lawyers should comply with any orders or rules requiring the disclosure of the use of GenAI. Maura R. Grossman, et al., Is Disclosure and Certification of the Use of Generative AI really Necessary?, 107 Judicature No. 2 (2023).

Erroneous analysis provided by GenAI, left unchecked, creates a closely related risk that a lawyer may end up pursuing unmeritorious claims on behalf of a client. Guidance at 4. Lawyers are prohibited from making claims or asserting defenses without probable cause, for the purpose of harassing or maliciously injuring any person. Rule 3.1(a)(1). Further, lawyers cannot present claims or defenses unwarranted by law. Rule 3.1(a)(2). Overreliance on GenAI to generate legal arguments could result in the presentation of an unmeritorious legal argument, for example, should GenAI provide fictional case law in support of a position that turns out not to exist.

How is a lawyer to bill for the use of GenAI? The Guidance provides that a lawyer may charge a client for the actual time spent in the use of GenAI, such as crafting GenAI prompts and editing GenAI output. Guidance at 4. But a lawyer cannot bill for the time saved by using GenAI. Guidance at 4. While some commentators have discussed whether the expense of GenAI should be absorbed in overhead, the Guidance provides that clients can be charged for the costs associated with GenAI, which the lawyer should explain in the fee agreement. Cal. Bus. & Prof. Code §§ 6147 (how costs will be handled), 6148 (bills should clearly identify the costs incurred and amount of costs). Clients can give informed consent to such charges through the fee agreement, but the overall billing of such charges in California is limited by unconscionability. Rule 1.5(a)-(b).

In contrast, the FBA Opinion would require disclosure of the intent to charge a client for GenAI and prohibit a lawyer from charging a pro-rated amount for period charges, which the lawyer advised should be handled as overhead. FBA Op. at 6.

Do you have to tell your client you are using GenAI? Lawyers must promptly inform clients of anything requiring client consent. Rule 1.4(a)(1). This would cover disclosure of confidential information to GenAI, for example. A lawyer must also reasonably consult with the client about the means by which to accomplish a client’s objectives in the representation and abide by a client’s decisions. Rule 1.4(b), 1.2(a). The Guidance, however, does not require disclosure of the use of GenAI to clients, but provides that a lawyer should evaluate their communication obligations throughout the representation, including the novelty of the technology, risks of GenAI, scope of representation and client’s sophistication. Guidance at 4. The lawyer should “consider disclosure” to the client regarding the use of GenAI, including how it will be used and the risks of such use, as well as the benefits.

The Guidance does not speak to lawyer advertising. GenAI can be involved in the practice of law for advertising and intake purposes. The FBA Opinion would caution against misleading chatbots and require lawyers to inform prospective clients they are communicating with an AI program rather than an employee or person. The Florida guidance also provides that a lawyer should consider screening questions regarding whether the inquiring person is already represented by another lawyer. FBA Op. at 7. The FBA Opinion would also allow lawyers to advertise their use of GenAI but would not allow lawyers to claim their use of GenAI is superior to other lawyers unless the claims are objectively verifiable, which is a factual question to be analyzed on an individual basis. FBA Op. at 7.

Missing Topics
The Guidance does not address several topics that have been discussed elsewhere, including whether GenAI engages in the unauthorized practice of law. Future ethics opinions issued by COPRAC may address nuances as GenAI technology evolves. For now, the Guidance lays out very helpful principles for the use of GenAI.

Carole J. Buckner is Of Counsel to Wilson Elser Moskowitz Edelson & Dicker LLP in the San Diego office. She is a longtime member of the OCBA’s Professionalism and Ethics Committee..