Brett J. Trout
ABA Formal Opinion 512
The American Bar Association has just released new Formal Opinion 512 covering the use of artificial intelligence (AI) for lawyers. Over the next decade AI will move from a risky endeavor to a necessary tool in the arsenal of all of the best attorneys. Given that your use of AI as an attorney is a matter of “when” rather than “if,” it is important to inform yourself now of the ethical rules concerning the unfathomably valuable resource that is AI. The fifteen-page opinion covers seven areas of concern: competence; confidentiality; communication; supervision, frivolousness; and billing.
Competence
The ABA states that while you do not have to become an AI expert to use AI in your practice, you must understand the benefits and risks of the particular AI system you intend to use. Alternatively, you may enlist the assistance of someone else who has such knowledge. Given the rapidly advancing nature of AI, you are also required to keep up to date on changes to the system you are using. The primary danger is AI’s known propensity to produce inaccurate, misleading, and/or outright false information. AI may even generate great quotes from fake case citations for cases that do not exist. It is therefore your job as an attorney to closely vet all AI-generated output you intend to use in your practice. The degree of vetting required will vary depending on how you intend to use the output. If you are only using the output to generate potential ideas, the AI output may not require any vetting at all. If you are using the AI-generated content to write a brief to be filed with the court, the output will require the highest form of vetting to ensure the brief is complete, accurate, and compelling. AI-generated content can be intoxicatingly on point, but if you cannot absolutely verify its accuracy, you are putting your clients and your practice at risk by using it. Failure to properly vet AI-generated content runs the risk of violating ABA Model Rules of Professional Conduct 1.1.
Confidentiality
The biggest ethical trap for lawyers using AI is the disclosure of confidential client information. While all lawyers are aware of the duty under ABA Model Rules of Professional Conduct 1.6 to maintain client confidences, many are unaware that when you input information into an AI system that system uses that input to train itself to generate future output. For instance, if you input a proposed confidential settlement agreement into an AI system and asked it to improve the agreement, the system would now have that initial agreement and be able to provide it to others asking the AI system to generate a settlement agreement. That output could easily lead to the disclosure of identifiable client confidences that could lead to you getting slapped with an ethical violation. Even if the AI system never publicly discloses any of the settlement agreement, the fact that it could, could itself put you on the wrong side of an ethics inquiry.
To address this issue it is possible to use a “closed” AI system. A closed AI system uses a trained AI system, but is not accessible by anyone but authorized users. Many law firms implement such systems to address this exact issue, ensuring that nothing input into the system is publicly accessible. It is still important to note, however, that even though the input in such systems is not publicly accessible, it is still available as output to other attorneys within the firm. For instance, if you input a brief with a novel legal argument into your firm’s internal AI-system, that same system may output that same novel legal argument to a colleague with a similar legal issue. You may then find your legal argument getting shut down in court before you even had the chance to flesh it out or argue it yourself. Accordingly, even with these types of internal AI systems great care must be taken to make sure that all output is vetted to remove any material that could possibly lead to the revelation of client confidences by unwitting colleagues. It is also incumbent upon any attorney using any AI system that the attorney be familiar with the security features, information retention policies, terms of use, privacy policy, and other contractual terms and policies governing use of the system.
Communication
To comport with ABA Model Rules of Professional Conduct 1.6 lawyers have a duty to disclose use of AI to clients. First, if a client ever asks about the lawyer’s use of AI, the lawyer must be completely forthcoming. Second, before ever inputting any client confidential information into an AI system of any type the lawyer must obtain prior, informed consent from the client acknowledging that the client understands the particular risks and benefits involved. Third, lawyers must also disclose to clients other uses of AI by the attorney depending on the scope of the representation, the client’s needs and expectations, the criticality of the use in the representation, and the sensitivity of the information. On one end, if you use AI to generate potential trial outcomes and specific damage awards, which the client will use to decide whether or not to file suit, you are obligated to reveal that use of AI to the client. On the other end, if you use AI to increase the efficiently of ordering office supplies that you use to perform general legal work, you are likely not required to disclose that use to your client. Regardless though of whether you are required to disclose a particular use of AI with your client, best practices dictate that the more accurately and comprehensively disclose your use of AI with your client the better and more productive the relationship will be for both you and your client.
Supervision
You are not only obligated to follow all of these rules. You are also obligated to ensure that all attorneys and non-attorneys in your firm under your supervision follow these rules as well.
Frivolousness
This section of the opinion reminds attorneys that courts have and will continue to reprimand attorneys who submit documents to the court that contain AI hallucinations, including fake cases, citations, and facts. If you do not vet all AI-output with a fine-tooth comb before submitting it to the court, it is only a matter of time before you find yourself reprimanded by the court and explaining to your state’s ethics board why you did not follow this advice.
Billing
If you bill by the hour and it takes you an average of ten hours to draft a brief you cannot charge ten hours if you use AI to cut your brief-writing time down to two hours. You cannot only bill hourly for the hours you actually worked. If you bill using flat fees or contingency arrangements you may be able to continue billing at your current rates even if AI significantly reduces the hours you spend on a particular project. No matter what though, any fee charged that involved little or no work on your behalf is an unreasonable fee. While you may charge a client directly for the cost of AI, there are several caveats. You must agree to such fees upfront, you cannot charge for AI that is wholly-integrated into the general services provided by the lawyer i.e. (AI grammar correction in word processing systems), you cannot charge more than the reasonable cost of the system to the attorney, and none of the AI charges may be duplicative of other charges. You also cannot charge clients for the time you spend getting up to speed with the use of general AI systems that you intend to use for other clients. If a client requests the use of a new specific AI tool which the lawyer must spend time learning and/or the AI tool has a per use cost, the lawyer may bill the client for those items provided the client has agreed upfront in writing.
Conclusion
The bottom line is that there should be no surprises in this new opinion. The opinion merely applies the existing rules to the current burgeoning technology. Stay apprised of the current ethical rules and the technology you are using in your practice and the application of the rules to the technology should not pose any monumental hurdles.
Recent Comments