> ## Content Index
> Fetch the complete content index at: https://www.realaigovernance.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Rule 3.3 Doesn't Care That a Chatbot Helped You "Paraphrase" a Quote
- URL: https://www.realaigovernance.com/rule-3-3-doesnt-care-that-a-chatbot-helped-you-paraphrase-a-quote/
- Published: 2026-08-22T13:00:45.000Z
- Updated: 2026-08-22T13:00:44.000Z
- Description: Rule 3.3 doesn't care that a chatbot helped you write the brief
- Author: Robert T. Boyer Ph.D.
- Tags: assessment sequencing authority, AI Governance, Legal Ethics, ai in law, Legal AI, aba formal opinion 512, rule 3.3, legal malpractice, professional responsibility, Generative AI, gen ai, attorney ethics, Law Firm Risk, legal tech, court sanctions, vendor due diligence

Rule 3.3 doesn't care that a chatbot helped you write the brief.

A federal court struck attorney Tyrone A. Blackburn's filing in *Cartagena v. Dixon* this July and referred him to the Grievance Committee. The problem: quotation marks around language that never appeared in the cases he cited. At least seventeen times. His defense was that he'd "paraphrased." He'd already paid a $5,000 sanction for the same habit once before.

No new rule was needed to reach that outcome. Candor toward the tribunal has been required since long before generative AI existed. A tool helping draft the sentence never touched that requirement.

Most firm AI policies miss this, and they don't miss it in one direction. They miss it in two, at the same time, in different departments.

Some firms read Formal Opinion 512 and concluded the ABA had effectively banned the technology, so the safe move is to stay away from it entirely. That's not what the opinion says. The Standing Committee's first formal guidance on generative AI, issued in July 2024, never tells a lawyer not to use it.

Other firms read the same opinion and concluded the opposite: that guidance from the ABA amounts to a blessing, and a quick AI-assisted draft clears whatever bar applies. This is the more expensive mistake. Opinion 512 doesn't lower the standard for AI-assisted work. It does something closer to the reverse: it takes duties that were already binding — competence, confidentiality, communication, candor, supervision of anyone using the tool on your behalf, reasonable fees — and confirms none of them move because a machine helped produce the output.

Here's the friction most summaries skip over. The opinion doesn't say how a firm should build the verification step that would have caught Blackburn's misquotes before a judge did. That part was always going to be left to firms to figure out on their own, and most haven't.

Twelve pages produced the opinion. A much shorter internal summary produced the confusion. Somewhere between the two, both the ban that was never there and the duties that were never lifted got lost.

If your firm's AI policy compresses Formal Opinion 512 into a sentence, that sentence is someone's interpretation, not the opinion. Go read Rules 1.1, 1.4, 1.6, 3.3, 5.1, 5.3, and 1.5 directly, and ask which one your last AI-assisted filing was actually checked against.

Blackburn's chatbot didn't write a false citation. He did, when he decided "paraphrase" was close enough. The verification step that would have caught it isn't a technology problem, and it's not one Opinion 512 was ever going to solve for you.

The Verification Protocol starts exactly there — candor and supervision, the two duties this case actually turned on.