Artificial intelligence joined legal practice with the advent of autocorrect, text editors, electronic discovery, search tools, and chatbots. Now it supplements most lawyers’ toolboxes, even if they are unaware of it, through pre-filling forms, identifying potentially related case law and suggesting relevant searches, or providing analyses of judges’ and opposing counsels’ decision-making. But that is the world of classic artificial intelligence.
A new era of artificial intelligence is rocketing into legal tech. For lawyers, this version of artificial intelligence creates new outputs (briefs, contracts, etc.) based on predictions derived from what it learns through databases it’s trained on. Counsel, courts, and bar associations are grappling with the impact of this “generative” artificial intelligence (“GAI”) on the practice of law and each lawyer’s duty of basic technological competence. GAI takes a giant leap forward yet presents significant ethical considerations for lawyers before and while using GAI in client representation.
What is GAI? If you’ve heard of ChatGPT, you’ve heard of GAI. At its most basic, GAI “learns” by being trained on a database (called a “large language model” or “LLM”) and then predicts the answer to questions (aka “prompts”) it is asked. As we briefly referenced in our August 8, 2024 Top Tip, the abilities of GAI lead to very important questions regarding compliance with a lawyer’s ethics obligations, particularly in the areas of competence, confidentiality, communication, consent, and fees.
This article dives further and deeper into the ABA Standing Committee on Ethics and Professional Responsibility’s Formal Opinion 512, Generative Artificial Intelligence Tools. Opinion 512 discusses the interaction of seven ethical duties and GAI, and this article will follow a similar outline while weaving in Texas rules where appropriate. All quotations in this article are from Opinion 512 unless otherwise noted.
Competence
Lawyers must provide competent representation to their clients. Model Rule 1.1 and TDRPC 1.01 both refer to “competence” as requiring the legal knowledge, skill, training, and preparation necessary for the representation. Both rules require a reasonable level of technological competence but do not require an expert level ability to use GAI technology. Instead, Opinion 512 notes “lawyers must have a reasonable understanding of the capabilities and limitations of the specific GAI technology that they might use.” To obtain and maintain competence, Opinion 512 suggests: “lawyers should consider reading about GAI tools targeted at the legal profession, attending relevant continuing legal education programs, and . . . consulting others who are proficient in GAI technology.”
Although some initially viewed GAI as a tool that should never be used in a client’s representation because of the inherent risks, the duty to be competent in technology suggests that lawyers at minimum need to become aware of the available GAI tools so they can evaluate whether GAI technology may allow them to serve their clients more efficiently while also increasing the quality of their work. In undertaking this evaluation, however, the duty of competence also requires lawyers to recognize the risks of GAI, including the risk its output is not accurate or that it could be hallucinating (“providing ostensibly plausible responses that have no basis in fact or reality”). These inaccuracies can lead a lawyer into providing inaccurate legal advice to clients or misleading information to courts or third parties, all of which violate the lawyer’s obligation to provide competent representation. For this reason, if a lawyer determines that GAI should be used to assist with a client’s matter, then independent verification or review of its output by a human is required.
The best way to view GAI is as “a springboard or foundation for legal work.” It is a starting place, not the end result. GAI should not replace a lawyer’s exercise of her professional judgment either. “Competent” representation presumes lawyers will use the necessary level of skill and judgment for all work because lawyers are fully responsible for all of the work performed for the client. This means lawyers must resist the temptation to have GAI advise the client, negotiate a claim, or otherwise usurp the lawyer’s role, even if for the honorable purpose of reducing the costs of the representation. As with all technology, the duty of competent representation requires lawyers to make informed decisions whether or not to use a particular GAI technology, and in so doing, the lawyer must keep her client’s interests and objectives at the forefront.
Confidentiality
Generative artificial intelligence presents unique confidentiality challenges, arising primarily from the potential exposure of confidential information either through a publicly accessible large language model (e.g. GPT) or through unauthorized access or inadvertent disclosure. Model Rule 1.6 and TDRPC 1.05 address a lawyer’s duty to keep confidential all information related to the representation of a client. To comply, Opinion 512 reiterates the guidance found in Comment 18 to Model Rule 1.6: “In considering whether information relating to any representation is adequately protected, lawyers must assess the likelihood of disclosure and unauthorized access, the sensitivity of the information, the difficulty of implementing safeguards, and the extent to which safeguards negatively impact the lawyer’s ability to represent the client.”
At least two confidentiality-related risks arise when using GAI. First, entering confidential information into public GAI sources (e.g. ChatGPT) can expose that information to any number of third parties in direct violation of the confidentiality obligation. Second, lawyers may use an “in house” GAI–one that trains and learns only on information provided within the firm–but they must still consider the risk that a client’s confidential information input into an “in house” GAI could be accessed by an authorized user who doesn’t know that the information is confidential or who could fail to adequately protect the information. For example, lawyers working on one client’s project could input confidential information into the “in house” GAI that is later utilized in response to prompts from other firm lawyers who then repeat it in court filings without knowing that it was a different client’s confidential information.
Many of the current GAI tools include these risks of direct or indirect disclosure, so a lawyer is required to obtain the clients’ informed consent, confirmed in writing prior to inputting information relating to the representation into the GAI tool. But beware, Opinion 512 suggests that “merely adding general, boiler-plate provisions to engagement letters purporting to authorize the lawyer to use GAI is not sufficient.” This is likely true in Texas because under the new terminology in the TDRPC, for consent to be informed, the lawyer must provide “adequate information and explanation about material risks of and reasonably available alternatives to the proposed course of conduct.”
To be able to provide that level of information to a client about the risks of possible disclosure, Opinion 512 encourages lawyers at a minimum to “read and understand the Terms of Use, privacy policy, and related contractual terms and policies of any GAI tool they use to learn who has access to the information that the lawyer inputs into the tool or consult with a colleague or external expert who has read and analyzed those terms and policies” and suggests consulting with IT professionals to better understand these risks. Then, having obtained that knowledge, the lawyer is required to give to the client her “best judgment about why the GAI tool is being used, the extent of and specific information about the risk, including particulars about the kinds of client information that will be disclosed, the ways in which others may use the information against the client’s interests, and a clear explanation of the GAI tool’s benefits to the representation.”
Communication
Even when the risks of disclosure don’t require informed consent, confirmed in writing, lawyers should still consider whether their obligations to keep clients reasonably informed about the representation mandates disclosure regarding the use of GAI in the representation. Model Rule 1.4 and TDRPC 1.03 obligate lawyers to explain to, and consult with, clients about the representation. Whether to disclose, and the scope of any required disclosure, are fact specific, but Opinion 512 offers some general guidelines about when disclosure of the lawyer’s use of GAI in the representation is required:
- When the client asks;
- When the terms of the engagement letter require it;
- When the terms of the client’s outside counsel guidelines require it (assuming that arrangement is relevant to nature of the representation);
- When information related to the representation could be entered into the GAI (as discussed above);
- When the use of a GAI tool is relevant to the basis or reasonableness of the fee;
- When the GAI output could influence the course of the representation; or
- When the use of GAI may violate the client’s reasonable expectations that the lawyer is applying her unique skill and judgment for which she was hired.
As TLIE suggested in our August 8, 2024 Top Tip, and as Opinion 512 agrees, erring on the side of disclosure is the better course of action. The engagement letter is a good place to communicate these disclosures to the client, and obtain informed written consent, if required.
Claims, Contentions, and Candor
Texas lawyers have well-established duties to assert only non-frivolous claims and contentions, and to act with candor toward the courts. Opinion 512 identifies several issues “that have arisen to date with lawyers’ use of GAI outputs” which “include citations to nonexistent opinions, inaccurate analysis of authority, and [the] use of misleading arguments.” In response, many courts require affirmative disclosure of the use of GAI or a written acknowledgement that any GAI generated results have been reviewed and confirmed by the filing party’s attorney. Lawyers must be aware when they use GAI that it can hallucinate or otherwise provide inaccurate information, incorrect statements of the law, false citations to authority, and other misleading information even though the output may sound accurate or plausible to the lawyer. TLIE recommends that each firm endorse a policy that requires verification by a human of any GAI output before that output is used in any manner before a tribunal or in the client’s legal matter.
Supervisory Responsibilities
Texas recently revamped TDRPC 5.01 regarding the duties of supervisory and managerial attorneys and imposed an obligation on these attorneys to implement policies and procedures that ensure compliance with the TDRPC. Those policies and procedures should also include the firm’s permissible use(s) of GAI and ensure that each lawyer and nonlawyer using GAI in the firm complies with their professional responsibilities regarding the use of GAI. Opinion 512 advises that policies and procedures should include adequate training “in the ethical and practical use of the GAI tools relevant to [the] work as well as on risks associated with relevant GAI use.” Furthermore, lawyers’ supervisory obligation extends to third party GAI providers and tools, even though they are outside the firm.
Fees
A not as obvious but still important consideration regarding the use of GAI is how attorneys bill for work product prepared with the assistance of GAI. Model Rule 1.5 and TDRPC 1.04 require fees to be reasonable, which means that lawyers who bill by the hour may earn a lower fee when GAI reduces the time spent for the particular task at hand. If using GAI reduces the time required from many hours to one hour, the billed time must represent the amount of time the lawyer actually worked on the client’s matter, not what he was able to bill previously on similar projects that didn’t include the use of GAI. Opinion 512 provides a helpful example for hourly-based work: “If a lawyer uses a GAI tool to draft a pleading and expends 15 minutes to input the relevant information into the GAI program, the lawyer may charge for the 15 minutes as well as for the time the lawyer expends to review the resulting draft for accuracy and completeness.” On the flat fee side, the lawyer should closely consider whether the flat fee is one that becomes unreasonable with the use of GAI if it results in little to no work being performed by the attorney.
Charging GAI work back to the client as an expense is also fraught with pitfalls and generally should be a pass-through cost, without a markup for the lawyer. Moreover, if the lawyer treats the GAI charges he incurs as office overhead, it cannot be passed through to the client without an agreement otherwise made prior to the work being performed.
Regarding technology as an overhead expense, Opinion 512 advises:
To the extent a particular tool or service functions similarly to equipping and maintaining a legal practice, a lawyer should consider its cost to be overhead and not charge the client for its cost absent a contrary disclosure to the client in advance. For example, when a lawyer uses a GAI tool embedded in or added to the lawyer’s word processing software to check grammar in documents the lawyer drafts, the cost of the tool should be considered to be overhead. In contrast, when a lawyer uses a third-party provider’s GAI service to review thousands of voluminous contracts for a particular client and the provider charges the lawyer for using the tool on a per-use basis, it would ordinarily be reasonable for the lawyer to bill the client as an expense for the actual out-of-pocket expense incurred for using that tool.
Lawyers must also remember that they cannot charge a client for increased time caused by the lawyer’s own inexperience. A lawyer should not learn GAI on the client’s dime, even if the GAI is used in the representation, unless the charges are agreed to upfront after informed consent, confirmed in writing, or when the client has requested the use of a particular GAI tool in the representation.
Has the State Bar of Texas Weighed In?
Not yet. But, the State Bar of Texas “(SBOT”) created a Task Force for Responsible AI in the Law (“TRAIL”), and this past summer, TRAIL provided 20 recommendations to the SBOT Board of Directors. The SBOT Board of Directors referred the recommendations to its Administration Committee for review and potential recommendations to the board. TRAIL also submitted a request to the Professional Ethics Committee for an ethics opinion seeking guidance on applying the TDRPC to the use of AI, including the lawyer’s duties addressed in the topics above.
Conclusion
As Opinion 512 recognizes, the speed at which GAI changes and develops means lawyers must be ever vigilant in their efforts to both utilize GAI and comply with the ethics rules. TLIE anticipates that the Professional Ethics Committee will release an opinion that aligns with Opinion 512, and TLIE will update its constituents when that opinion becomes available. In the meantime, TLIE encourages lawyers to thoroughly review Opinion 512, to become educated on GAI’s uses and risks, and to err on the side of caution by communicating clearly with clients and disclosing GAI’s use clearly and upfront. TLIE also offers forms for its insureds regarding policies and procedures a firm can implement for the responsible use of GAI, as well as form disclosures that can be given to the client to obtain informed written consent from the client regarding the firm’s usage of GAI. Insureds can email rskmgmt@tlie.org to request copies of these forms.