Update on AI Ethics for Lawyers

Updates regarding previously authored article by Peggy Reetz (Partner).

The American Bar Association Standing Committee on Ethics and Professional Responsibility (“the Committee”) released a formal opinion in July 2024 covering the use of Artificial Intelligence (AI) in the practice of law (specifically, “generative artificial intelligence”). As discussed in an earlier edition of this publication, early uses of AI tools by lawyers produced “mixed” results in terms of ethical compliance and as an effective tool for advocacy (what client wants to pay for AI generated gobbledygook that not only does not advance their cause but results in sanctions?) (link). The potential for reliance on such tools has come under scrutiny across the globe.  See, “International Lawyers Facing AI and Legal Malpractice: Navigating Innovation, Obligation, and Bar Regulations Disputes,” Ibrahim Ati, November 18, 2024 (link). So, it is timely that the ABA is providing additional guidance for lawyers facing the demands and challenges in utilizing AI. 

In Formal Opinion 512, the Committee notes that lawyers using generative AI tools must fully consider applicable ethical obligations (link).  Fair enough, most lawyers are generally guided not to fall outside any ethical obligations, including duties to provide competent legal representation, protecting privilege and confidentiality while advancing meritorious claims with candor before a tribunal, and oh yes, charging reasonable fees (maybe a whole book chapter should be devoted to parsing out which part of a bill should be allocated to non-AI work as compared to AI-reliant activities).  As a starting point, the opinion outlines that lawyers should be mindful of several model rules: competence (Model Rule 1.1); confidentiality of information (Model Rule 1.6); communications (Model Rule 1.4).  Lawyers typically seek to achieve these separate and apart from whatever tool gets them there.  So, how should these be applied in the context of using AI tools?   

The opinion provides illustrations of some practical considerations: 

Competence:

• Lawyers must have a reasonable understanding of the capabilities and limitations of the technology
  • Tools that increase efficiency may contain inherent risks (including so-called “hallucinations” where an ostensibly plausible AI generated response has no basis in fact or reality) 
  • Independent verification may be required  
  • New tools may lead to new competency expectations 

Confidentiality:

• Reasonable efforts should be made to prevent inadvertent, unauthorized disclosure of, or access to information relating to client representation; certain AI tools may pose a greater risk of improper disclosure
  • Internal access and disclosure protections to prevent inadvertent disclosure 
  • Some AI tools commingle client and lawyer generated information 
  • Re-using AI generated content may inadvertently include client data  
  • Informed consent required (boilerplate provisions not sufficient) 
  • Which information will be disclosed to outside parties given the type of tool utilized?

Communication:

• Client communications are built from obligations to act as a fiduciary (prompt dissemination of information, meeting client objectives, advice for informed decision-making, client’s expectations for information consistent with duty to act in client’s best interests and requirements)
  • Which part of using a generative AI tool requires obtaining informed consent? 
  • Express disclosures may be required where terms of engagement dictate such conditions 
  • Will use of AI influence a significant decision in the representation or is reliance on such a tool key to an outcome (e.g., jury selection)?

Meritorious Claims/Candor to Tribunal:

• AI generated content cannot be false, deceitful or misrepresentative  
• Courts have started to require disclosure of use of AI

Supervisory Responsibilities: 

• Subordinates, assistants must be supervised and trained on AI uses 
• Vendor policies and credentials must be verified 
• Tools must be configured to preserve confidentiality and security; reliability and limitations of liability to be examined 
• Vulnerabilities identified

Fees: 

• Flat fees – if AI was primarily the source of the work-product, is the fee reasonable? 
• AI charges as an expense – disclosure required (typically would be viewed as overhead) 
• No fees for learning to use the technology

The opinion concludes by noting the evolving nature of generative AI and how this in and of itself presents challenges to meet all the above ethical and competency obligations.  The catch-all for these considerations is “vigilance.”   

Author: Peggy Reetz is a partner at Mendes & Mount, LLP and a member of the TIPS Cybersecurity and Data Privacy Committee. 

Original article published by the American Bar Association Tort Trial and Insurance Practice Section. Permission to republish granted by the American Bar Association.