Duty of Competence: Understanding AI Before Using It

Rule 1.1 requires lawyers to provide competent representation. This now includes understanding AI tools. You cannot use generative AI for legal work without knowing how it works. You must understand its capabilities. You must understand its limitations. You must understand when it might fail.

The ABA compares AI to calculators and legal databases. These are tools that enhance competence. But they also require knowledge to use properly. A calculator gives wrong answers if you press wrong buttons. AI gives wrong answers if you prompt poorly or accept outputs without review.

Lawyers cannot delegate legal judgment to AI. You can use AI for research and drafting. But you must review the output. You must verify the facts. You must check the citations. You must apply your professional judgment. The AI suggests. You decide.

Continuing legal education now includes AI competence. Some states require AI CLE credits. Others will follow. You must stay current on AI developments. You must understand new models and new risks. Competence is an ongoing requirement, not a one-time certification.

Confidentiality in the AI Era (Rule 1.6)

Client confidentiality remains absolute. Rule 1.6 requires lawyers to protect client information. This includes information entered into AI systems. You cannot paste client details into consumer AI tools that store data. You cannot use AI that trains on your inputs. These actions breach confidentiality.

Enterprise AI systems offer protections. They often sign Business Associate Agreements. They do not train on client data. They offer SOC 2 certification. They provide audit logs. But you must verify these protections. You cannot assume. You must review the terms of service. You must review the privacy policy.

Sub-processor agreements are critical. AI vendors often use third-party services. These sub-processors might handle your client data. You must know who they are. You must know their security practices. You must ensure they are bound by confidentiality obligations. The chain of protection must remain unbroken.

Inadvertent disclosure is still disclosure. If you paste client information into ChatGPT and it appears in another user’s output, you have breached confidentiality. This is malpractice. This is an ethics violation. This is potentially a crime. The risk is real. The consequences are severe.

Candor to Court: Verification and Disclosure

Rule 3.3 requires candor to the tribunal. This includes verifying all factual and legal representations. When AI generates content, you must verify it. The Mata v. Avianca case provides the warning. Lawyers cited fake cases generated by ChatGPT. The judge sanctioned them. The bar investigated them. Their careers suffered permanent damage.

Courts increasingly require AI disclosure. Some local rules mandate certification that AI-generated content was verified. Others require disclosure of AI use in briefs. You must know the rules for every court where you file. You must comply with certification requirements. Failure to disclose is itself a violation.

Human review is mandatory, not optional. You must read every case cited by AI. You must pull the full text. You must verify the holding. You must verify the facts. You cannot rely on AI summaries. You cannot rely on AI quotes. You must go to the source.

If you discover an AI error after filing, you must correct it immediately. Rule 3.3 requires you to correct false statements. The duty of candor continues after filing. If you learn that AI-generated content was wrong, you must notify the court. You must withdraw the erroneous material. Delay compounds the violation.

Billing Ethics: Charging for AI-Assisted Work

Rule 1.5 requires reasonable fees. This includes billing transparency. If AI reduces the time needed for a task, you cannot bill the old hourly estimate. You cannot charge for three hours of research when AI found the answer in ten minutes. You must bill for the actual work performed.

Florida takes the strictest approach. The state requires disclosure of AI use when it affects fees. If AI saves time, you must pass those savings to the client. You cannot pocket the efficiency gains as pure profit. The fee must reflect the value delivered, not the time saved by technology.

Fee agreements need updating. Your retainer agreement should address AI use. It should specify how you bill for AI-assisted work. It should disclose any AI costs you pass through. Clients deserve transparency. They deserve to know how technology affects their bills.

Value-based billing may replace hourly billing for AI-assisted work. If AI drafts a motion in one hour that previously took ten, billing one hour might be unreasonable. The value to the client is the same. The fee should reflect value, not time. This shift requires new billing models and client communication.

The Growing Patchwork of State Rules

California requires disclosure for “significant” AI use. This includes drafting substantive legal documents, analyzing legal issues, or conducting research that informs legal advice. The state also requires lawyers to explain AI use to clients in plain language. You cannot bury disclosure in fine print.

New York counties vary in their requirements. Some mandate disclosure for AI use in submissions to the court. Others require certification that AI-generated content was verified for accuracy. You must check local rules before filing in New York state courts.

Federal courts are developing their own standards. Some magistrate judges require AI disclosure in discovery responses. Others require it in briefs. The federal judiciary is watching the state developments. Federal rules may follow the state patchwork or establish uniform standards.

Multi-state practitioners face the greatest complexity. A lawyer licensed in three states must comply with three different sets of AI rules. The most restrictive state may govern your practice. You must know the overlap. You must comply with the strictest standard.

The Myth vs The Reality

Common Questions About AI and Legal Ethics

Brief Summary

AI-generated legal content is ethical only when lawyers maintain human oversight and comply with evolving bar requirements. ABA Formal Opinion 512 establishes baseline competence, confidentiality, and candor standards. State rules vary dramatically, with California requiring 9 key considerations and Florida mandating billing disclosure. Lawyers must verify all AI outputs, protect client confidentiality in AI systems, and comply with court-specific disclosure rules. Billing must reflect actual work performed, not time saved by AI. The most significant risk remains unverified citations, as demonstrated by the Mata v. Avianca sanctions. No AI tool is “bar certified” – compliance depends entirely on lawyer conduct, not vendor claims.