Ethics, Professional Responsibility, and the AI Era
Deliberate Academy Editorial Team
Reviewed for accuracy and professional relevance
You're 6 lessons in — don't lose your progress.
Sign up free to save where you are and earn a verified certificate when you pass.
- Explain why the duty of competence, duty of candour, and fiduciary duty to clients are each directly implicated by AI use in legal practice
- Apply the team-member framing — treating AI as fast, broadly knowledgeable, occasionally confidently wrong, and without professional standing — to determine the correct supervisory standard for AI outputs
- Describe the current SRA, BSB, Law Society, and judicial guidance on AI use in legal practice and identify the consistent direction of regulatory travel
- Build an AI governance log that demonstrates reasonable diligence in the event of a regulatory inquiry
Legal professionals operate under a professional responsibility framework that has not fundamentally changed because AI tools now exist. The duty of competence still requires that advice is based on accurate legal analysis. The duty of candour still prohibits misleading the court. The fiduciary duty to clients still requires that their interests are placed above the lawyer's convenience. What has changed is the environment in which those duties are exercised — and the new failure modes that AI introduces if those duties are exercised without adequate care.
The Duties That Remain Non-Delegable
Competence. The professional duty of competence — the obligation to provide legal services to the standard of a competent practitioner — cannot be delegated to an AI tool. A lawyer who produces an advice letter based on AI-generated legal analysis that has not been verified for accuracy has not met the competence standard, regardless of how sophisticated the tool. The SRA's competence framework requires that solicitors maintain sufficient understanding of the law and process relevant to their area of practice. Using AI in a way that substitutes for that understanding rather than augmenting it is an incompetent practice, not an efficient one. In the United States, ABA Model Rule 1.1 (competence) imposes the equivalent standard, requiring the legal knowledge, skill, thoroughness, and preparation reasonably necessary for the representation — a standard that ABA and state bar guidance increasingly interprets as including a reasonable understanding of the AI tools a lawyer relies on and their failure modes.
Candour to the court. The duty of candour — the absolute prohibition on knowingly misleading the court — is the most fundamental obligation in litigation practice. Submitting a document to a court that contains a fabricated case citation, even if the lawyer did not know it was fabricated because they did not check, raises serious candour questions. The courts that have imposed sanctions in AI citation cases have not accepted ignorance of the AI's error as a complete defense. The duty of candour requires verification. The lawyer knew, or should have known, that AI outputs require verification.
Fiduciary duty. In contexts where a fiduciary relationship exists between lawyer and client — most obviously in solicitor-client relationships — the lawyer must act in the client's best interests. Using AI tools in a way that is convenient for the lawyer but potentially harmful to the client (by introducing confidentiality risks, by reducing the rigor of advice, or by presenting AI-generated output as qualified professional advice without adequate review) raises fiduciary duty concerns.
The courts and professional regulators are developing their approach to AI-related professional conduct issues in real time. SRA enforcement activity, court sanctions for AI citation failures, and regulatory guidance from the Law Society and BSB are all evolving. What is tolerated today may be sanctioned next year as awareness of AI failure modes increases and regulatory expectations harden. Building rigorous AI practice disciplines now — verification, supervision, data governance, client disclosure — is not merely current compliance. It is future-proofing against a regulatory environment that is tightening.
How AI Does and Does Not Change the Professional Responsibility Framework
AI does not create new professional duties. It creates new contexts in which existing duties must be applied. The duty to verify the accuracy of legal advice has always existed — AI makes it more important to discharge that duty explicitly rather than assuming the tool has done it. The duty to protect client confidentiality has always covered third-party disclosures — AI tools create a new category of third-party recipient that must be assessed against that duty. The duty to supervise fee-earner work has always applied in firms — AI-generated work requires the same supervision as any other output produced by someone other than the reviewing lawyer.
The frame that helps is this: AI is a member of your team who is extremely fast, very broadly knowledgeable, occasionally confidently wrong, and has no professional standing. You would not send a client advice letter written entirely by an unqualified colleague without reviewing it. You would not submit a court filing drafted by a paralegal without checking every citation. The professional responsibility structure for AI output is the same — the tool's lack of professional status makes your supervisory role more important, not less.
This framing is not merely a teaching device — it tracks how US professional guidance has approached the question directly. ABA Model Rule 5.3, which governs a lawyer's responsibilities for the conduct of nonlawyer assistance (traditionally paralegals and support staff working under the lawyer's direction), has been applied by analogy to AI tools in ABA and state bar ethics guidance: the supervising lawyer must make reasonable efforts to ensure that a nonlawyer's work — human or AI — is compatible with the lawyer's professional obligations, and remains responsible for that work product. The rule was not written with AI in mind, but its logic transfers directly onto the team-member framing above.
A senior associate argues that AI tools should be held to a lower standard of supervision than junior fee-earners because AI is a tool rather than a person, and tools are not subject to professional responsibility. How does the lesson's framework address this argument?
Select one answer.
Emerging Regulatory and Professional Body Guidance
The professional body landscape on AI is developing rapidly. Key developments as of 2026:
The SRA has confirmed that its competence and supervision requirements apply to AI-assisted work and has issued specific guidance on confidentiality risks from AI tool use. It has indicated that it will consider AI governance as part of firm supervision assessments going forward.
The Bar Standards Board has confirmed applicability of existing duties and is monitoring developments in AI use in court proceedings, particularly following high-profile citation incidents. Guidance on disclosure of AI use in court documents is under active consideration.
The Law Society has published a practice note on AI in legal services covering supervision, data protection, client disclosure, and liability. It has established a working group on AI governance in legal practice.
The Judiciary in England and Wales has issued guidance on the use of AI in preparing documents for court proceedings, requiring disclosure when AI has been used in document preparation and emphasizing that the responsible lawyer must verify all content.
In the United States, the American Bar Association has issued ethics guidance addressing generative AI use in legal practice, and a growing number of individual state bar associations have followed with their own guidance or ethics opinions interpreting how existing competence, confidentiality, and supervision rules apply to AI tools. As in England and Wales, the direction of travel is toward more jurisdictions publishing guidance and toward that guidance hardening over time, not toward AI use being treated as a special case exempt from ordinary professional responsibility rules.
Build an AI governance log for your practice: document which AI tools you use, what you use them for, what data protection assessment you have conducted for each, and what review process you apply to AI-generated outputs before they go to clients or courts. Review it every six months against current regulatory guidance. This log demonstrates the reasonable diligence that is the difference between a regulatory inquiry that is resolved and one that becomes a conduct finding.
AI Governance Log Under Regulatory Inquiry — Mid-Size Private Client Firm
Context
A partner at a mid-size private client firm received an SRA enquiry following a client complaint about the quality of estate planning advice. The complaint was unconnected to AI, but during the inquiry the SRA asked the firm to describe its governance arrangements for AI tools used in matter work. The firm had been using an AI drafting tool for Wills and trust documents for several months, but had no documented governance policy and had not formally assessed whether the tool's data handling terms were compatible with its confidentiality obligations.
Action
The partner was unable to produce documentation of the firm's AI governance approach during the inquiry. The SRA indicated it would expect firms to be able to demonstrate their supervision and governance arrangements for AI-assisted work as part of its standard oversight responsibilities. Following the inquiry, the partner established an AI governance log covering all tools used across the department: the tool name, its approved use cases, the data protection assessment conducted, the review process applied to AI outputs before they went to clients, and the date of last review. She also commissioned a data protection assessment of the AI drafting tool's vendor terms.
Outcome
The original client complaint was resolved separately. The SRA inquiry did not produce a conduct finding, in part because the partner engaged constructively and committed to implementing governance documentation. The partner noted that the absence of a log had meant she could not demonstrate what her team had or had not done — and that a conduct inquiry where you cannot reconstruct your own processes is a significantly more exposed position than one where you have a documented record. The governance log was reviewed at the firm's next partnership meeting and adopted across all practice groups.
Building an AI-Responsible Practice
The legal professionals who will build sustainable, high-performing practices with AI are not those who use the most AI — they are those who use AI most responsibly. Responsible AI use in legal practice means:
Using AI to do more genuine legal work, not to cut corners on the verification and judgment that legal work requires. Using AI tools that are fit for purpose and appropriately governed, not the most convenient consumer tool available. Maintaining the professional skills — legal analysis, research methodology, drafting judgment, client relationship — that AI augments rather than replaces. Being transparent with clients about AI use when that transparency serves their interests or their right to informed consent.
The enduring advantage of a legal professional in an AI-enabled world is not speed — it is professional judgment, relational trust, and accountable expertise. AI makes competent lawyers more productive. It does not make it possible to bypass competence.
A barrister uses AI to draft a skeleton argument for a complex commercial arbitration. The skeleton is well-structured and cites several cases. The barrister reviews it for logical flow but does not verify the citations in primary sources before filing. The arbitral tribunal later identifies that two of the cited cases are fabricated. What professional duties are most directly engaged?
Select one answer.
Exercise
Your Task
Start your AI governance log today. Create a simple document with four columns: AI tool used, what you used it for, data protection assessment (is there a signed DPA, is client data excluded, what is the vendor's training-use policy), and the review process you applied to the output before it went to a client or court. Populate it with your current AI tool usage over the past month, filling in any assessment gaps by checking the relevant vendor terms. Set a calendar reminder to review and update the log every six months. This 10 to 15 minute setup exercise creates the documented evidence of reasonable diligence that the lesson identifies as the difference between a regulatory inquiry that resolves and one that becomes a conduct finding.
Your reflection
Did you complete this exercise? What did you find? (Saved locally in your browser)
- The duty of competence, duty of candour, and fiduciary duty to clients are non-delegable to AI tools — the lawyer remains professionally responsible for every output produced under their supervision, regardless of how it was generated.
- Treating AI as a team member who is extremely fast, very broadly knowledgeable, occasionally confidently wrong, and has no professional standing is the practical frame that correctly locates the supervisory responsibility in legal AI use.
- Regulatory guidance from the SRA, BSB, Law Society, and Judiciary is developing rapidly and consistently in the direction of tighter disclosure requirements and clearer professional accountability for AI-assisted outputs — building rigorous practice disciplines now is future-proofing, not over-compliance.
- An AI governance log — documenting tools used, purposes, data protection assessments, and review processes — demonstrates reasonable diligence and is the difference between a regulatory inquiry that resolves and one that becomes a conduct finding.
- The enduring advantage of a legal professional in an AI-enabled world is professional judgment, relational trust, and accountable expertise — AI makes competent lawyers more productive, but cannot substitute for the competence itself.