Navigating the New Frontier: State Bar Guidance on Generative AI for Trusts and Estates Counsel

The advent of Generative Artificial Intelligence (GAI) is rapidly transforming the legal landscape, including for trusts and estates lawyers. From drafting initial documents to legal research, the potential efficiencies are tantalizing. However, this powerful new tool brings with it a host of ethical considerations that practitioners must navigate with care. Recognizing this, the State Bar of California’s Standing Committee on Professional Responsibility and Conduct in November 2023 issued “Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law,” providing a crucial framework for attorneys by applying existing professional obligations to this new technology.¹ This article examines this guidance by exploring the key ethical rules implicated by GAI, with a special focus on how the pervasive problem of AI “hallucinations” can lead to violations.

I. Duty of Competence (Rule 1.1)

The duty of competence now explicitly includes keeping abreast of “the benefits and risks associated with relevant technology” pursuant to Rule 1.1, Comment [1] of the Rules of Professional Conduct. The primary risk associated with GAI is its potential to “hallucinate”—producing outputs that are false or inaccurate.² These inaccuracies can range from “false statements,” such as citing fabricated case law, to more subtle “false assertions,” like misstating the holding of a real case.³

For an estate planner, this risk is profound. In a hypothetical from the State Bar’s training materials, an attorney, “Larry,” uses a GAI tool to generate a will, but fails to verify the AI’s application of state-specific estate laws and client-specific wishes. This overreliance leads to significant errors in the distribution of assets, contravening the client’s wishes and creating grounds for a will contest.⁴ Larry’s ethical failure was not in using AI, but in his failure to competently supervise the technology by critically reviewing and validating its output.

While it may be tempting to assume that this risk is confined to general-purpose GAI products like ChatGPT or Gemini, the State Bar’s materials make clear this is not the case. Even law-specific tools from major providers like LexisNexis and Thomson Reuters, which are developed for lawyers and often use Retrieval-Augmented Generation (RAG) technology trained on a curated set of authoritative legal materials, are susceptible to hallucinations.⁵ A May 2024 Stanford study highlighted in the State Bar’s training found that these specialized legal research tools still produce false or misleading information. The study focused on Lexis+AI, Westlaw AI-Assisted Research, and Ask Practical Law, and found that each hallucinated between 17 and 33 percent of the time.⁶ These errors were not simple mistakes; they included complex legal misunderstandings, such as misinterpreting court holdings and failing to distinguish between a party’s argument and the court’s actual ruling.⁷ This underscores that a lawyer’s duty to independently verify AI-generated work is absolute, regardless of the sophistication of the tool. The duty of competence requires more than just detecting false results; it demands that a lawyer’s professional judgment is never delegated to the AI.⁸

II. Candor to the Tribunal (Rule 3.3) and Meritorious Claims (Rule 3.1)

The risk of hallucinations extends directly to our duties before the court. An attorney who submits a probate petition, a brief in a trust contest, or any other court filing containing hallucinated case law or a misstated legal standard violates their duty of candor to the tribunal. Similarly, filing a claim or contention based on a hallucinated legal theory that lacks basis in law or fact would violate the duty to assert only meritorious claims. Before any GAI-assisted document is submitted to a court, the attorney has an absolute obligation to review all outputs for accuracy and correct any errors or misleading statements.⁹

III. Duty of Confidentiality (Bus. & Prof. Code, § 6068(e); Rules 1.6, 1.8.2)

While hallucinations relate to the output of GAI, the input process implicates our duty of confidentiality. Many GAI products use the information that is input—including prompts and uploaded documents—to train the AI model, which may lack adequate security or share the data with third parties. Therefore, an attorney must not input any confidential client information into a GAI tool without ensuring the platform has adequate confidentiality and security protections.¹⁰

IV. Communication with Clients (Rules 1.2, 1.4)

An attorney’s reliance on a hallucinated analysis can directly breach the duty of communication.¹¹ If an estate plan is structured around an AI’s erroneous interpretation of a tax law, or if a beneficiary is advised on their rights based on a hallucinated understanding of trust distribution rules, the attorney has failed to provide the client with accurate information needed to make informed decisions about the representation. California’s guidance advises attorneys to evaluate their communication obligations based on the facts and circumstances, including client sophistication and the risks of the specific tool. ABA Formal Opinion 512, though only advisory in California, states that informed written consent should include the lawyer’s judgment on why the tool is being used, the specific risks of disclosure, and the tool’s benefits to the representation.¹²

V. Fees and Billing (Rule 1.5; Bus. & Prof. Code, §§ 6147-6148)

An attorney may charge for the actual time spent productively using GAI, such as crafting effective prompts and verifying the results. However, an attorney must not charge hourly fees for time saved by using GAI. It could also be improper to bill a client for time spent wrestling with and correcting an AI’s fabricated information, as this could result in an unconscionable or unreasonable fee.¹³

The advisory ABA Formal Opinion 512 offers a helpful distinction for practitioners, suggesting that if a GAI tool functions like general overhead (e.g., Grammarly), its cost should not be charged to the client, whereas per-use charges for a specific GAI service may be reasonably billed as an expense.¹⁴

VI. Supervisory Duties (Rules 5.1, 5.2, 5.3)

Managerial and supervisory lawyers must establish clear policies on the permissible uses of GAI and provide training on its ethical pitfalls, including the risk of hallucinations. They must make reasonable efforts to ensure all firm personnel comply with their professional obligations when using these tools.¹⁵

Conclusion

Generative AI offers powerful new capabilities, but its use is fraught with ethical risks. For the estate planning and probate bar, where precision and trust are paramount, these risks demand heightened vigilance. The State Bar’s guidance makes clear that GAI must be a tool to augment, not replace, our professional judgment and obligations.

Download the original publication (PDF) →