AI technologies have the potential to transform the practice of law by enhancing the efficiency and quality of legal work. On one level, AI systems are simply tools- no different from email services or word processes. On another, they present unique risks. While they are simply ‘tools’, there are some unique characteristics that need careful consideration: including whether confidential data is used to ‘train’ an AI model, and whether lawyers are able to continue to exercise professional judgment.
As AI becomes more prevalent, lawyers will need to develop competence with these advanced technologies to deliver the best possible service to their clients, and we believe to satisfy their duty to clients under professional conduct rules. They must understand how AI tools work, evaluate their strengths and limitations, and properly supervise the results.
Some argue that overreliance on AI could compromise a lawyer’s professional independence or judgment. If lawyers follow AI recommendations without question or are unable to properly evaluate AI outputs due to a lack of technical skill, they may not be exercising independent judgment as required by conduct rules. However, when properly supervised by lawyers with competence in the technology, AI can enhance legal judgment rather than replace it.
Lawyers must critically evaluate AI results in the same way they would work from junior colleagues and be capable of reaching their own conclusions. With diligent oversight and an understanding of its capabilities, AI can support lawyer independence rather than undermine it. We talk more about supervision of AI systems below.
Knowledge and Skill
Most professional conduct rules impose a duty of competence, requiring lawyers to have proper knowledge and skill in the law and legal practice. In some jurisdictions understanding technology is specifically acknowledged as a professional responsibility.
Comment 8 to the ABA Model Rules notes, this duty extends to “the benefits and risks associated with relevant technology.” Lawyers must keep abreast of AI tools, get proper training, and understand how to evaluate AI outputs. In Victoria, a Practice Note issued by the Supreme Court of Victoria states that “the use of common technologies is a core skill for lawyers and a basic component of all legal practice.” Pursuant to the Uniform Rules, a solicitor must provide clear and timely advice to assist a client to understand relevant legal issues and to make informed choices about action to be taken during the course of a matter, consistent with the terms of the engagement. AI systems can assist a lawyer in both ensuring that advice is clear and timely in that it can act as a quality control measure, can improve the clarity of advice, and can do so very quickly.
The need for lawyers to understand and properly apply AI is clear, AI can (and will get better at) improve outcomes and reduce the risk of overcharging and providing a substandard service. However, lawyers must also understand the limitations and risks of any AI tool to properly advise clients on its use and oversee its role. This requires ‘supervision.’
‘Supervision’ means exercising professional judgement in the critical review of the output of an AI tool. Supervision can be challenging where the quality of output appears to be excellent. The quality of output of systems such as Titan often exceeds the quality of work performed by people. Supervision requires knowledge of the areas in which AI systems can hallucinate and the adoption of strategies to reduce this risk. With close supervision, lawyers can take advantage of AI’s benefits while still satisfying their professional duties.
Disclosure of client and confidential information
AI tools present issues for lawyers’ duties to keep client information confidential and privileged. Many AI technologies require large volumes of data to function properly, including potentially confidential client information. However, lawyers must still maintain client confidences.
The obligations of confidence are very clear. In those states in Australia that have adopted the Uniform Rules, for example, a solicitor must not disclose any information which is confidential to a client…except as permitted in Rule 9.2. Rule 9.2.1 – A solicitor may disclose information which is confidential to a client if the client expressly or impliedly authorises disclosure. If a lawyer uses AI and the client’s confidential information is used to help improve the AI’s capabilities, the lawyer would need the client’s consent to disclose that information, in accordance with Rules 9 and 9.2.1. Titan does not use any user data (uploads) or outputs to retrain itself or any other model.
When using AI, lawyers should clarify what data is being used and take appropriate precautions. If an AI tool needs to analyze confidential documents or communications to function, lawyers must consider if they need client consent, require the data be kept confidential, used only for the specific task, restrict access to the information, and have the data permanently deleted afterwards. Close oversight and strong security practices are essential to using AI ethically while safeguarding client confidences.
Recommended Next Steps
Below are some recommended next steps for lawyers who want to understand and apply AI in their work.
Education and Training:
- Pursue comprehensive training on AI’s capabilities, limitations, and functionalities.
- Engage in continuous learning due to the rapid technological advancements.
Exercise Due Diligence:
- Regularly evaluate the accuracy and reliability of AI outputs.
- Treat AI results as you would any legal advice—always question, verify, and ensure it aligns with legal best practices.
Ensure Independent Judgement:
- Do not rely solely on AI recommendations; always use personal judgment to maintain professional independence.
Supervise AI Outputs:
- Implement protocols for AI system oversight and critically analyze AI results, understanding potential pitfalls.
Client Consent and Confidentiality in AI Retraining:
- Be transparent with clients about how their data might be used by AI, especially for AI retraining.
- Secure explicit client consent if their confidential data will be used to enhance the AI’s capabilities.
This article was inspired and took lessons from this article published in the Law Society Journal.