The Ethics of Artificial Intelligence Prepared by David L. Gordon Office of the General Counsel Jackson Lewis P.C. (404) 586-1845 | GordonD@jacksonlewis.com Rebecca L. Ambrose Office of the General Counsel Jackson Lewis P.C. (518) 512-8723 | firstname.lastname@example.org This paper is meant to provide information of a general and educational nature and does not constitute legal advice or create an attorney-client relationship. Readers should consult counsel of their own choosing to discuss how these matters relate to their individual circumstances. The views expressed in this paper are solely those of the authors and do not necessarily represent the views of their firm or its principals or clients. Reproduction in whole or in part is prohibited without the express written consent of Jackson Lewis. ©2017 Jackson Lewis P.C. 1 The use of artificial intelligence (“AI”) raises a variety of complicated ethics issues for attorneys. Part I of this paper will review recent changes in technology and how ethics committees across the country have sought to apply the Rules of Professional Conduct in the evolving digital world, often resulting in inconsistent opinions across, and sometimes within, jurisdictions. Part II of this paper will review various ways in which AI is being utilized in the legal profession. Part III of this paper will discuss particular Rules of Professional Conduct that apply to lawyers using AI. This paper supplements a presentation on this topic delivered at the 2017 Jackson Lewis Corporate Counsel Conference. I. APPLICATION OF THE RULES OF PROFESSIONAL CONDUCT IN THE EVOLVING DIGITAL WORLD In the not so distant past, before the advent of email, blackberries, Wi-Fi and social media, application of the Rules of Professional Conduct to the conduct of lawyers was, in many respects, straightforward. However, as technology has evolved and its use has become common place in the legal profession, ethics committees around the country have struggled to apply those rules to situations involving technological advancements. The result has been the publication of ethics opinions in various jurisdictions (and sometimes within the same jurisdiction) related to technology that are often inconsistent. One such example involves the ethics opinions issued by various committees relating to lawyers’ social media use. Several ethics committees have considered the extent to which a lawyer can send a “friend” request on social media to an unrepresented witness in an effort to engage in informal discovery related to a pending litigation. In a 2010 ethics opinion issued by the New York City Bar’s Committee on Professional Ethics, that committee concluded that a lawyer governed by the New York Rules (or the lawyer’s agent) can ethically send a “friend request” from a legitimate personal social media account to an unrepresented person’s social networking page with the intent of obtaining information from that page for use in connection with litigation without disclosing the reason for the request.1 However, ethics committees in other jurisdictions have reached a different conclusion, suggesting that in order to ethically “friend” an unrepresented witness, the lawyer would need to disclose information regarding the reason for the friend request at the time it is sent.2 Ethics committees also struggle with the use of LinkedIn. In March of 2015, the New York County Lawyers Association Professional Ethics Committee issued an opinion which concluded that while it is ethically permissible for lawyers to maintain a LinkedIn profile, profiles that contain more than basic biographical information likely constitute attorney advertising and therefore have to comply with the applicable advertising rules, including those rules requiring the use of certain disclaimers.3 Shortly after that opinion was issued, the New York City Bar’s Committee on Professional Ethics considered the same question and, applying the same Rules of Professional Conduct, reached a different conclusion.4 That committee concluded that there may be several reasons why a lawyer might use LinkedIn, and a lawyer’s LinkedIn profile (or other LinkedIn content) would only constitute attorney 1 Ass’n of the Bar of the City of N.Y. Committee on Prof’l Ethics, Formal Op. 2010-2 (2010). 2 See, e.g., Philadelphia Bar Association Professional Guidance Committee, Op. 2009-02 (March 2009); San Diego County Bar Ass’n, Legal Ethics Op. 2011-2 (May 2011). 3 N.Y. County Lawyers Ass’n Professional Ethics Committee, Formal Op. 748 (March 10, 2015). 4 Ass’n of the Bar of the City of N.Y. Committee on Prof’l Ethics, Formal Op. 2015-7 (2015). 2 advertising if it met certain specified criteria.5 As such, in New York, there are currently inconsistent ethics opinions regarding a lawyer’s use of LinkedIn, making compliance difficult. Another thorny area is metadata, specifically, when it comes to a lawyer’s ethical obligation to scrub metadata and whether lawyers are permitted to mine for metadata.6 While most jurisdictions have concluded that lawyers have an affirmative obligation under their duty of confidentiality to their clients to remove potentially privileged or confidential metadata from documents prior to sending them to third parties7 , ethics committees have come to vastly different conclusions on whether lawyers can ethically mine for metadata in the documents that they receive from an adversary.8 Recognizing the challenges faced by lawyers in applying the Rules of Professional Conduct to situations involving technological advancements, the ABA Commission on Ethics 20/20 engaged in a three-year study regarding the impact that globalization and technology are having on the practice of law and how the regulation of lawyers should be changed to address those developments. As a result of that study, in 2012 the ABA Commission on Ethics 20/20 recommended a number of changes to the ABA Model Rules of Professional Conduct and accompanying commentary to address those changes.9 Several of those amendments were ultimately adopted by the ABA House of Delegates and have a direct impact on the conduct of lawyers in using technology, social media and AI. These amendments will be addressed in further detail in Part III of this paper. II. WHAT IS AI AND HOW IS IT BEING USED IN THE PRACTICE OF LAW? The appropriate application of the Rules of Professional Conduct to a lawyer’s use of social media and other technology is sometimes difficult to determine, and the ethics issues become even murkier when lawyers are using more advanced technology, such as AI. While these considerations will be explored in more detail in Part III below, before discussing the applicable Rules of Professional Conduct, it is important to understand exactly what AI is and how it is being used in the legal industry today. 5 Id. The criteria set forth by the N.Y. City Bar ethics committee for determining whether a LinkedIn profile constitutes attorney advertising are as follows: “(a) it is a communication made by or on behalf of the lawyer; (b) the primary purpose of the LinkedIn content is to attract new clients to retain the lawyer for pecuniary gain; (c) the LinkedIn content relates to the legal services offered by the lawyer; (d) the LinkedIn content is intended to be viewed by potential new clients; and (e) the LinkedIn content does not fall within any recognized exception to the definition of attorney advertising.” 6 For purposes of this paper, the contemplated metadata is that which may be included in documents exchanged between adverse parties and counsel in a matter, not metadata which might be embedded in documents that are subject to discovery. 7 See, e.g., N.Y. State Bar Ass’n Committee on Prof’l Ethics, Op. 782 (Dec. 2004); see also, D.C. Bar Legal Ethics Committee, Op. 341 (Oct. 2007); Pennsylvania Bar Ass’n Committee on Legal Ethics and Professional Responsibility, Formal Op. 2009-100 (2009). 8 Compare N.Y. State Bar Ass’n Committee on Prof’l Ethics, Op. 749 (Dec. 2001) (“in light of the strong public policy in favor of preserving confidentiality as the foundation of the lawyer-client relationship, use of technology to surreptitiously obtain information that may be protected by the attorney-client privilege, the work product doctrine or that may otherwise constitute a ‘secret’ of another lawyer’s client would violate the letter and spirit of these Disciplinary Rules”) and Washington State Bar Ass’n Rules of Professional Conduct Committee, Advisory Op. 2216 (2012) (concluding generally that lawyers may mine for metadata so long as “special software” is not used to recover metadata that is not easily accessible). 9 ABA Commission on Ethics 20/20 Introduction and Overview (April 6, 2017), available at http://www.americanbar.org/content/dam/aba/administrative/ethics_2020/20120508_ethics_20_20_final_hod_introdution_ and_overview_report.authcheckdam.pdf. 3 Artificial intelligence is defined as “the capability of a machine to imitate intelligent human behavior.”10 One area where artificial intelligence is increasingly being used is in electronic discovery - through a process called predictive coding. Predictive coding “uses a type of AI programming that allows the computer…to learn from attorney instruction.”11 Very generally, a machine engages in an interactive process with an attorney through which it learns from the attorney how to classify documents in a particular case (e.g., as relevant or irrelevant, privileged or not privileged).12 The machine is able to extrapolate input provided by the attorney on a small sampling of documents to “classify the complete collection” and “rank the probability of each document fitting into the classification.”13 While the use of predictive coding likely will still require a second level manual review by attorneys, limiting the universe of documents that will require a human set of eyes can result in substantial cost savings and can help decrease the likelihood of human error during the first level of review. A second area in which we have seen the use of AI in the legal profession is through the use of certain document assembly software. 14 Software that is used to prepare tailored templates of standard form documents has been used by law firms to increase productivity and decrease the likelihood of human error that normally exists in the preparation of such documents. Another area in which we see AI being used is in software such as Lex Machina and Ravel Law, which are programs that are able to review court decisions and other data to make predictions about judges and lawyers, such as giving the probability that a particular motion will be approved by a specific judge or whether a particular lawyer typically settles cases or brings them to trial.15 Finally, AI is revamping the way that lawyers can conduct legal research. A good example is IBM’s Ross, which was built on IBM’s Watson technology.16 Ross has recently been “hired” by several law firms to get lawyers up to speed on relevant law, and was specifically built to be able to summarize the status of the law in a an understandable way.17 Ross can also keep attorneys up to date on developments in the law that could impact their matters.18 10 Artificial Intelligence, MERRIAM-WEBSTER (April 6, 2017), available at https://www.merriamwebster.com/dictionary/artificial%20intelligence. 11 Ralph C. Losey, A Survey of Emerging Issues in Electronic Discovery: Predictive Coding and the Proportionality Doctrine: A Marriage Made in Big Data, 26 REGENT UNIVERSITY U.L. REV. 7, 21 (2013-2014). 12 Id. at 21 – 24. 13 Id. at 21. 14 This paper is limited to a discussion of document assembly software that is used by lawyers and law firms. It does not extend to potential issues faced by non-lawyers who are using and/or marketing such products. 15 Steve Lohr, A.I. Is Doing Legal Work. But It Won’t Replace Lawyers, Yet., N.Y. TIMES (Mar. 19, 2017), available at https://www.nytimes.com/2017/03/19/technology/lawyers-artificial-intelligence.html?register=email&auth=registeremail&_r=3. 16 Cecille De Jesus, AI Lawyer “Ross” Has Been Hired By Its First Official Law Firm, N.Y. TIMES (April 6, 2017), available at https://www.nytimes.com/2017/03/19/technology/lawyers-artificial-intelligence.html?register=email&auth=registeremail&_r=3. 17 Id. 18 Id. 4 III. APPLICATION OF THE RULES OF PROFESSIONAL CONDUCT TO THE USE OF ARTIFICIAL INTELLIGENCE IN THE PRACTICE OF LAW As set forth in more detail below, there are several Rules of Professional Conduct that are directly implicated by a lawyer’s use of AI. Duty of Competence Under Rule 1.1 of the ABA Model Rules of Professional Conduct (“Model Rules”)19, lawyers are required to provide competent representation to their clients. That rule states that “[c]ompetent representation requires the legal knowledge, skill, thoroughness and preparation reasonably necessary for the representation.”20 While Model Rule 1.1 has arguably always required a lawyer to keep current on changes in the law, including the impact that the use of technology may have on their representation of their clients, in 2012 the ABA adopted a change to the comments to Rule 1.1 which expressly states as much. Comment  to Model Rule 1.1 now specifically states that “[t]o maintain the requisite knowledge and skill, a lawyers should keep abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology….”21 To date, at least 27 states have adopted this amendment to the comments to the version of Rule 1.1 that is in place in their jurisdictions.22 In 2015, the State Bar of California Standing Committee on Professional Responsibility and Conduct issued Formal Opinion 2015-193, in which it considered the ethical obligations of a lawyer in handling discovery of electronically stored information. That committee concluded that a lawyer’s obligations under California’s rules governing competence “evolve as new technologies develop and become integrated with the practice of law.”23 Specifically, that committee stated that competence in the practice of litigation “requires, among other things, and at a minimum, a basic understanding of, and facility with, issues related to e-discovery, including the discovery of electronically stored information.”24 The committee concluded that “[a]n attorney lacking the required competence for ediscovery issues has three options: (1) acquire sufficient learning and skill before performance is required; (2) associate with or consult technical consultants or competent counsel; or (3) decline the representation.”25 As the California ethics opinion demonstrates, in order to fulfill their obligations to their clients under Rule 1.1, lawyers have an affirmative ethical obligation to understand the technology that is being used currently in the legal profession and the risks and benefits to the use of such technology. While competence in this area can be attained by associating with the appropriate consultants and counsel, 19 For purposes of this paper, we refer to the ABA Model Rules. However, whenever a lawyer is faced with an ethics issue, it is important that they review the Rules of Professional Conduct applicable to their conduct. 20 MODEL RULES OF PROF’L CONDUCT R. 1.1 (2016). 21 MODEL RULES OF PROF’L CONDUCT R. 1.1 cmt. (2016). 22 Robert Ambrogi, 13 15 17 18 20 21 23 24 25 26 27 States Have Adopted Ethical Duty of Technology Competence (April 7, 2017), http://www.lawsitesblog.com/2015/03/11-states-have-adopted-ethical-duty-of-technology-competence.html. 23 The State Bar of California Standing Committee on Professional Responsibility and Conduct, Formal Op. 2015-193 (2015). 24 Id. 25 Id. 5 failing to maintain the requisite level of competence related to technology can result in a violation of the rules, as well as other potential negative ramifications. Duty to Communicate A rule that works hand-in-hand with Model Rule 1.1 is Model Rule 1.4, which governs a lawyer’s duty to communicate with clients. Model Rule 1.4(a)(2) requires a lawyer to “reasonably consult with the client about the means by which the client’s objectives are to be accomplished.”26 If a lawyer intends to use AI in the provision of legal services, he or she has an obligation under Model Rule 1.4 to discuss that decision with clients. That discussion should include a conversation about the risks and limitations of any AI that the lawyer is contemplating using. Similarly, if for whatever reason a lawyer chooses not to use AI or technology in a particular matter in which it may be beneficial to the client to do so, a lawyer arguably has an obligation under Model Rule 1.4 to discuss that decision with the client as well. Duty of Confidentiality The duty of confidentiality is the bedrock of the attorney-client relationship. Under Model Rule 1.6, lawyers are strictly prohibited from revealing a client’s confidential information unless the client consents to the disclosure, the disclosure is impliedly authorized to carry out the client’s representation or one of the very narrow exceptions to the duty of confidentiality applies. 27 Additionally, under Model Rule 1.6(c), lawyers are required to “make reasonable efforts to prevent the inadvertent or unauthorized disclosure of, or unauthorized access to, information relating to the representation of a client.”28 Many of the AI programs that lawyers are using to provide their clients with legal services are supplied by third-party vendors who are not lawyers or otherwise associated with a law firm. As a result, lawyers must take steps, as required by Model Rule 1.6, to ensure that their clients’ information is appropriately safeguarded. To ensure the confidentiality of the information, a lawyer contemplating the use of AI should consider discussing confidentiality concerns with the third-party provider and should inquire about, among other things, what type of information is going to be provided, how that information will be stored, what security measures are in place with respect to the storage of the information, and who is going to have access to the information. Only after concluding that the client confidential information that they provide will be reasonably safeguarded should a lawyer proceed with using AI in connection with the representation of a client.29 26 MODEL RULES OF PROF’L CONDUCT R. 1.4 (2016). 27 MODEL RULES OF PROF’L CONDUCT R. 1.6 (2016). The confidentiality obligations imposed on lawyers with respect to current clients are extended to former clients as well under Model Rule 1.9(c). 28 MODEL RULES OF PROF’L CONDUCT R. 1.6 (2016). Comment  to Model Rule 1.6 notes that the “[f]actors to be considered in determining the reasonableness of the lawyer’s efforts include, but are not limited to, the sensitivity of the information, the likelihood of disclosure if additional safeguards are not employed, the cost of employing additional safeguards, the difficulty of implementing the safeguards, and the extent to which the safeguards adversely affect the lawyer’s ability to represent clients (e.g., by making a device or important piece of software excessively difficult to use).” 29 Model Rule 1.6 also works directly with Model Rule 1.1 when it comes to a lawyer’s use of AI. If a lawyer fails to possess or attain the requisite level of competence when it comes to understanding the technology that they are using, it could lead to an inadvertent disclosure of client confidential information, which could violate Model Rule 1.6 See, e.g., California Standing Committee on Professional Responsibility and Conduct Formal Op. 2015-193. 6 Duty to Supervise Under Model Rules 5.1 and 5.3, lawyers have an express obligation to supervise the work of both the lawyers and nonlawyers that they engage to assist them in providing legal services to ensure that their conduct complies with the Rules of Professional Conduct. While these rules are typically applied to humans, a 2012 adoption of an amendment to Model Rule 5.3 makes clear that these rules likely extend to AI as well. In 2012, the ABA approved the Ethics 20/20 Commission’s recommendation to change the title of Rule 5.3 from “Responsibilities Regarding Nonlawyer Assistants” to “Responsibilities Regarding Nonlawyer Assistance.”30 This change shows that the rule is intended to have reach beyond human assistants, to other nonlawyers, human or not, involved in the representation of a client. Under Rules 5.1 and 5.3, lawyers who engage AI to assist in the provision of legal services need to take steps to ensure that they are adequately supervising the AI’s work. How best to do this will depend on the type of AI being used, but compliance with these rules will likely require the lawyer using AI to take steps to verify that the AI’s work product is complete and correct, and doesn’t otherwise expose the client to risk by, for example, inadvertently disclosing client confidential information. IV. CONCLUSION Technological advances will continue to change the way that lawyers practice law. We already have seen AI enter the legal profession through lawyers’ use of predictive coding to increase the efficiency and decrease the cost of document review. We also have seen AI play an increasing role in the way that lawyers perform legal research and analyze court decisions. While it may take time for the black letter text of the Rules of Professional Conduct to evolve with these changes, it is important for lawyers to be cognizant of and compliant with their current ethical obligations with respect to the use of such technology. 4837-4008-1478, v. 1 30 See supra at fn. 10 at 12.