Module 5.3 · Topic 5
Professional Responsibility for Agent-Assisted Work
Bottom Line Up Front: Delegation does not transfer responsibility. Supervise agent actions, maintain competence, document participation and review, retain human judgment where law requires it. Inadequate supervision…
5.1 Supervisory Duties When AI Takes Action
ABA Model Rule 5.1 requires you to supervise others and ensure compliance with rules. This extends to agents. You remain accountable for agent decisions and compliance.
Supervisory duty requires: (1) Monitor agent actions for rule compliance. If an agent makes client, discovery, or strategy decisions, review for confidentiality and candor rule breaches. (2) Establish clear guidelines about agent autonomy versus decisions requiring approval. Rule 5.3 makes you responsible for nonlawyer (including AI) compliance. (3) Maintain competence in the work the agent performs. Rule 1.1 (Competence) makes clear: competence remains your responsibility, even with AI tools.
5.2 Accountability and Attribution for Agent Output
Agent-produced work bears your professional stamp. Mata v. Avianca: attorney submitted AI-generated fake citations to court. Sanctions followed. The attorney, not the agent, faced consequences.
Malpractice Risk:
Inadequate agent-work verification: Delegation of substantive work (contracts, research, client comms) without personal review creates malpractice exposure. Courts expect verification and documented review. Delegation without verification is evidence of negligence.
5.3 Documentation Requirements for Agent-Assisted Decisions
Courts now mandate disclosure of AI use in filings. Documentation protects you by showing instructions provided, output reviewed, facts verified, and judgments exercised.
- Document the task: Save prompt/spec, date, agent type, scope.
- Document review: Record what you verified, facts checked, reasoning red-teamed, comparisons made.
- Document judgment: Note your decisions: accepted, rejected, modified, or escalated agent recommendations.
- Disclose AI use: In filings, contracts, or client comms where material, disclose. WA Bar requires disclosure to clients when material.
- Retain records: Keep all documentation in your file. Defends against malpractice claims.
5.4 Ethical Frameworks for Autonomous AI in Legal Practice
Professional responsibility rules were written before AI agents existed. Mapping agent-assisted workflows onto existing ethics rules requires judgment about where human decision-making must be retained and where agents can operate autonomously. ABA Formal Opinion 512 (The Paradigm for Generative AI in Legal Practice) provides the clearest current guidance.
| Task Type | Applicable Rule | Agent Autonomy vs. Human Judgment |
|---|---|---|
| Legal research | Rule 1.1; 3.3 | Agent researches; human verifies citations before court submission. No exceptions. |
| Client communication | Rule 1.6; 1.4 | Agent drafts routine notices. Cannot send legal advice or confidential info without review. |
| Contract review | Rule 1.1 | Agent drafts templates; human reviews substance, compliance, risk. Agent review alone insufficient. |
| Due diligence | Rule 1.1; 1.6 | Agent summarizes and flags issues. Human reviews materiality and risk. Agent output is starting point only. |
| Litigation strategy | Rule 1.1; 3.3; 3.4 | Agent researches precedents. Human makes strategy decisions. Agent cannot draft motions for filing without review. |
| Conflict checks | Rule 1.7; 1.8 | Agent collects intake and flags conflicts. Human makes final determinations. Mandatory human judgment. |
Critical principle: Agents cannot fulfill duties requiring human judgment or discretion. Rule 1.6 (Confidentiality) — agent cannot solely guard secrets. Rule 3.3 (Candor) — you ensure honesty with courts. NC Bar Opinion 1 makes clear: AI does not relieve ethical duties. (See Module 7.1 on bar standards and Module 7.3 on risk management.)