ABA Opinion 512 explained: what solo lawyers actually need to do

Most explainers on Formal Opinion 512 read as if they were written for a two-hundred-lawyer firm with a general counsel, an IT security team, and outside ethics counsel on retainer. You don't have any of that. You have a caseload, a laptop, and the same six duties the ABA says apply to everyone.

The Standing Committee on Ethics and Professional Responsibility issued Formal Opinion 512 on 29 July 2024. By mid-2026 it's the document every state bar cites when a lawyer gets caught pasting client facts into a chatbot, or filing a brief with a case that doesn't exist. The opinion doesn't invent new rules. It takes six you already know and applies them to generative AI. What most write-ups skip is what each one actually asks of a solo practitioner, as opposed to a firm with a department for that.

The six duties, in plain terms

Model Rule 1.1, competence, requires a reasonable understanding of an AI tool's capabilities and limitations before you rely on it, kept current as the tool changes. Rule 1.6, confidentiality, requires assessing the risk that a tool discloses or reuses client information, and getting informed consent when that risk is real. Rule 1.4, communication, requires telling clients enough about your AI use for them to make informed decisions about the representation. Rule 3.3, candor toward the tribunal, means verifying every citation and fact an AI tool hands you before it goes near a court. Rules 5.1 and 5.3, supervision, apply even when the only person you supervise is yourself, or one paralegal. Rule 1.5, fees, limits what you can bill for time an AI tool saved you.

The competence problem nobody solves for you

A large firm handles Rule 1.1 with a vendor security questionnaire, a pilot period, and a memo from IT. You don't have that pipeline, and the opinion doesn't lower the bar because you're solo. "Reasonable understanding" still means reading the tool's documentation, testing it on a matter where nothing is at stake, and knowing what it does when it doesn't know an answer. Most generative AI tools guess rather than admit uncertainty. That is the one piece of competence you cannot skip: assume any specific citation, quote, or fact needs independent verification, every time.

Confidentiality: what to check before you paste anything in

The confidentiality duty comes down to one habit: read the data retention terms before the first time you use a tool on a client matter, not after. Free consumer tiers of most major chatbots train on your inputs by default. Paid business or enterprise tiers usually don't, but "usually" is not a defense in a bar complaint. Our confidentiality checklist breaks down what ChatGPT, Claude, Copilot, and Gemini actually retain by tier. If you haven't checked your current tools against it, that's a better use of twenty minutes than reading the rest of this article.

The billing trap in Rule 1.5

Opinion 512 says that if AI lets you draft a pleading in fifteen minutes instead of two hours, you can only bill for the fifteen minutes plus your review time, if you bill hourly. Flat fees get more latitude, but not unlimited latitude: the opinion still expects the fee to be reasonable for the work actually done. Thomson Reuters Institute's analysis flagged the logical trap here directly: taken too literally, the opinion's reasoning rewards a lawyer who works slower over one who uses better tools, and defines reasonableness by how long a task took rather than the value of the outcome. Until the rule catches up, the safe move for solo practice is straightforward. Negotiate flat fees where you can, and if you bill hourly, tell the client upfront that you use AI tools and that your fee reflects judgment and review time, not just the clock.

Candor and supervision aren't theoretical anymore

Courts stopped treating hallucinated citations as an amusing anecdote sometime around Q1 2026. US courts imposed at least $145,000 in sanctions for fabricated AI content in that quarter alone. Oregon set the tone in December 2025, when a Portland attorney was sanctioned $2,000 under a new per-infraction schedule: $500 for each fabricated citation, and $1,000 for a fabricated quotation, a formula federal courts in the same jurisdiction have since adopted. A separate 2026 review counted more than 500 documented instances of AI-hallucinated content reaching US court filings since early 2025. None of it required a finding of bad faith. A citation that doesn't exist is a Rule 3.3 problem the moment it's filed, regardless of intent.

For a solo lawyer, the Rule 5.1 and 5.3 supervision duties apply to yourself and to any AI tool you delegate drafting to, the same way they'd apply to a first-year associate. Read every citation. Open every case. If you don't have time to verify a brief before filing, you don't have time to use the shortcut that produced it.

What to actually do this week

Five habits cover most of what Opinion 512 requires, without needing a compliance department:

1. Read the retention and training terms for every AI tool before it gets anywhere near client information.
2. Keep a one-page log of which tools you use for which tasks, and update it when you add a new one.
3. Verify every citation, quote, and factual claim an AI tool generates before it reaches a client or a court.
4. Add a short AI-use paragraph to your engagement letter, covering your review practice and billing basis.
5. Revisit your fee structure for AI-assisted tasks. If you bill hourly for work AI meaningfully sped up, adjust it or disclose it.

If you don't have a written AI policy yet, that's the governance document this checklist belongs inside, and our AI policy template gets you a usable draft in an afternoon. For a longer look at exactly when the communication duty under Rule 1.4 requires telling a client you used AI, versus when it doesn't, see our piece on when disclosure is actually required.

Formal Opinion 512 will keep getting cited in disciplinary complaints for years. It was written assuming a firm with the infrastructure to build a vetting process around it. You don't need that infrastructure to comply. You need five habits, and the discipline to do them before the AI-assisted brief goes out the door, not after a judge asks where the case came from.

Get a ready-to-use engagement letter paragraph and a vendor-vetting checklist mapped to Opinion 512 in The Lawyer's AI Starter Kit, free. And if you're deciding which tools are actually safe enough to build a practice on, the best AI tools for lawyers page has the current DPA and retention status for each one.

FAQ