California State Bar Releases Guidance on Use of GenAI in Practice of Law

By: Kewa Jiang

On November 16, 2023, the California State Bar released the Practical Guidance For the Use of Generative Artificial Intelligence in the Practice of Law, which the Bar states should be considered “as guiding principles rather than as ‘best practices.” The guidance comes at a time of accelerated development and use of generative artificial intelligence (GenAI). The legal profession particularly has been the subject of articles for years that speculate the replacement of attorneys with AI, such as six years ago and this year. However, attorneys are adapting and using GenAI in their legal practice. Though, some attorneys’ experiments with GenAI are turning into cautionary tales resulting in sanctions and less than favorable outcomes for clients.

In the recently released guidance, the California State Bar acknowledges that GenAI has “wide-ranging applications for the practice of law and administrative functions of the legal practice” but states that GenAI “must be used in a manner that conforms to a lawyer’s professional responsibility obligations.” The California State Bar cautions lawyers to understand the risks and benefits of using GenAI, taking into consideration a range of factors such as the client, the matter, the practice area, the firm size, and the tools themselves, ranging from free and readily available to custom-built, proprietary formats.

Below is a highlight of some of the practical guidance provided by the Bar in relation to lawyers’ professional responsibilities.

• Duty of Confidentiality: The guidance warns against uploading or inputting confidential client information in GenAI platforms. The Bar notes that GenAI platforms may “lack reasonable or adequate security.” The guidance suggests lawyers should:
  • Consult with IT or cybersecurity professionals to ensure the AI system used “adheres to stringent security, confidentiality, and data retention protocols.”
  • Review the Terms of Use and other information to determine how inputted information is used.
  • Ensure the platform does not share inputted information with third parties.
  • Ensure the platform does not use the information for its own use, such as to train or improve its own product.

• Duties of Competence and Diligence: Given the instances of GenAI platforms “hallucinating” false or fictional cases and the potential embedded bias within GenAI, the guidance reminds lawyers of their core duties of competency and diligence. Lawyers have a duty to competently use technology, “including the associated benefits and risks,” and to “apply diligence and prudence with respect to facts and law.” It is not enough for an attorney to simply detect and eliminate false AI-generated results. Attorneys must engage in the “active practice of law and application of trained judgment.”
  • The guidance warns against overreliance of AI tools and the use of AI tools as the end all be all to legal research and analysis. Rather, a lawyer has a non-delegable duty of “professional judgment,” which remains “the lawyer’s responsibility at all times.” While AI may help supplement human-performed research or humans can supplement AI-performed research, traditional legal research and writing remain critical to attorney analysis.

• Charging for Work Produced by Generative AI and Generative AI Costs: The guidance states that a lawyer may use GenAI to “more efficiently create work product” and may charge for the time actually spent on the work product, such as refining AI inputs or prompts, or reviewing GenAI outputs. However, a lawyer “must not charge” the client the time saved due to the use of GenAI. The guidance recommends a fee agreement that explains the fees and cost of using GenAI.

Looking Ahead

In looking ahead, other state bars are also considering guidelines on the use of AI in the legal profession, such as New York and Florida. But, until the day AI replaces all attorneys, an attorney must still exercise their best judgment and heed the internal warning bells that ring when something isn’t right, whether the information is human or machine generated.