AI Legal Ethics: ABA Model Rules and Formal Opinion 512

ABA Formal Opinion 512 (July 2024) is the most consequential AI ethics document for U.S. attorneys. It clarifies how the ABA Model Rules apply when lawyers use generative AI. This is the practical operator read.

The short version: AI use is permitted, even encouraged, when handled competently. The lawyer remains accountable for the output. The supervisory and confidentiality obligations are not delegated.

The four Model Rules that matter most

Rule 1.1 (Competence) — A lawyer must provide competent representation, including understanding the benefits and risks of relevant technology. Generative AI is now considered relevant technology. Lawyers should understand:

• How AI tools work at a basic level
• The risks of hallucination, bias, and confidentiality breach
• When AI is appropriate for which tasks

You don't need to be a developer. You do need to understand what the tools do and don't do.

Rule 1.6 (Confidentiality) — Lawyers must not disclose client confidences. AI tools that retain client data, use it for training, or share it externally violate this rule. Practical implications:

• Free consumer ChatGPT, Claude, or other AI tools: do not use with client data
• Enterprise tiers with proper data handling: appropriate when configured correctly
• Document review platforms: must have proper data handling certifications

Rule 5.1 / 5.3 (Supervision) — Lawyers are responsible for the conduct of non-lawyer assistants. AI tools are treated like non-lawyer assistance:

• Supervising lawyers must ensure AI is used competently
• Junior attorneys should not unilaterally adopt AI for client work
• Verification of AI output is non-delegable
• Workflow design must include attorney review

Rule 7.1 (Communications) — Communications about a lawyer's services must not be false or misleading. AI-generated marketing materials are still subject to attorney advertising rules:

• AI-drafted marketing content requires attorney review
• Claims of expertise, success rates, or capabilities must be substantiated
• Compliance officers should review AI marketing workflows

What Formal Opinion 512 actually says

The opinion's core conclusions:

• Competence requires AI understanding. Lawyers who use AI must understand its capabilities and limitations sufficient to evaluate output.

• Confidentiality requires deliberate tool selection. Lawyers must select AI tools that protect client confidentiality. Free or consumer-tier AI typically does not.

• Supervisory obligations apply. Partners and supervising lawyers must oversee AI use by junior attorneys and staff just as they oversee non-lawyer assistants.

• Billing must be honest. Lawyers cannot bill clients for time AI completed in seconds at hourly rates as if the work took the historical time. Value-based billing and modified hourly approaches may be appropriate.

• Candor to tribunal applies. AI-generated citations and arguments must be verified. The Mata v. Avianca facts illustrate the consequence of skipping verification.

What this means in practice

At the individual lawyer level:

• Use enterprise-tier AI tools only for client work
• Verify every citation, quote, and legal proposition AI generates
• Disclose AI use to clients when appropriate (engagement letter)
• Bill honestly for AI-assisted work
• Document supervisory review

At the firm level:

• Adopt firm-wide AI policy under Model Rules
• Train attorneys on AI ethics and use
• Configure AI tools for confidentiality
• Document workflow design for AI-assisted work
• Annual compliance review

State-by-state developments

In addition to the ABA Model Rules, individual states have issued AI guidance:

• California: State Bar guidance on practical use issued 2023, updated 2025
• New York: Bar Association AI Task Force published guidance
• Florida: Specific opinion on AI use for legal research
• Texas: Bar guidance with focus on confidentiality
• Illinois: ARDC guidance addressing supervision
• Washington D.C.: Bar opinion on AI in advocacy

Check your state's specific rules. ABA Model Rules are advisory; state rules govern.

Honest billing for AI-assisted work

Formal Opinion 512 specifically addresses billing:

• Cannot bill 8 hours for what AI completed in 30 minutes
• Can bill for attorney time spent verifying, supervising, and refining
• Can use value-based fees that better reflect AI-augmented work
• Cannot bill hourly rates for "savings" passed back to clients

This is changing law firm economics. Firms billing aggressively on time AI saved face client pushback and ethics complaints. Firms adopting value-based or capped fees for AI-augmented work are more sustainable.

Engagement letter language

Many firms now include AI disclosure in engagement letters:

"In providing legal services, our firm may use artificial intelligence tools to assist with research, drafting, document review, and related tasks. All output is reviewed and verified by the attorneys representing you, and client confidentiality is maintained through tools that protect privileged information. Our billing reflects the value of our work, including the time spent reviewing and verifying AI-assisted output."

Tune to firm style. The disclosure protects client expectations and the firm.

Insurance and risk management

Legal malpractice insurers are starting to ask:

• Do you use AI in client work?
• What tools?
• What supervision is in place?
• What training have attorneys received?

Some insurers offer modest premium reductions for firms with documented AI policies and training. Some add exclusions for AI-related errors without verification. Read your policy.

What can go wrong

The patterns we see at firms that get into trouble:

Pattern 1: Unauthorized use. Associate uses consumer ChatGPT with client documents. Firm has no policy. Confidentiality is breached.

Pattern 2: Unverified citations. Mata v. Avianca pattern. Brief filed with hallucinated cases. Sanctions or worse.

Pattern 3: Honest billing failures. Firm bills hourly rates for AI-completed work at historical durations. Client discovers. Reputation and ethics consequences.

Pattern 4: Inadequate supervision. Junior lawyers run AI tools without partner review. Errors compound. Malpractice exposure.

Pattern 5: Marketing claims. AI-generated marketing makes unsupported claims. Bar discipline.

Each pattern is preventable with a structured AI policy and training. Each is increasingly common when firms deploy AI without ethics infrastructure.

What we recommend

For firms deploying AI:

• Written firm-wide AI policy under Model Rules
• Approved tool list (Tier 1: approved without conditions; Tier 2: approved with documented use; Tier 3: prohibited)
• Annual attorney training (90 minutes minimum, documented)
• Engagement letter language addressing AI use
• Workflow design with attorney verification gates
• Quarterly compliance review of AI usage
• Annual policy refresh

Most of this is one-time setup plus light ongoing maintenance. Firms that invest in this infrastructure operate AI confidently. Firms that don't operate AI with ongoing exposure.

Bottom line

AI ethics for attorneys is not a barrier to AI use — it's the operating manual. ABA Formal Opinion 512 makes the framework clear: competent use, confidentiality protection, supervisory oversight, honest billing, candor to tribunal, accurate communications.

Firms that build this infrastructure can deploy AI aggressively and ethically. Firms that ignore it deploy AI at growing legal and reputational risk. The competitive firms in 2027-2030 will be the ones that took the ethics work seriously today.