AI is in your practice whether you invited it or not. Your clients are using it. Your staff is using it. Opposing counsel is using it. And the courts are absolutely paying attention.
So here's the point of this post: AI can make you faster, sharper, and more profitable—or it can hand your credibility to the judge in a paper bag. The difference is whether you treat AI like a serious tool with real guardrails… or like a magic button.
Bar associations across the country are converging on one theme: AI doesn't change the rules—we still do. Competence. Confidentiality. Supervision. Candor. Fees. All of it.
AI is rising because the market is forcing it
If you're waiting for "certainty," you're going to be late.
The Thomson Reuters Institute has reported GenAI usage at work nearly doubling year-over-year (with legal professionals leading the pack), and majorities expecting AI to have a high or transformational impact in the next five years.
Clients are also showing up AI-informed. Clio's 2025 reporting points to more than half of consumers having used (or being willing to use) AI to answer a legal question, and a meaningful chunk being directed to contact a lawyer. Translation: clients will increasingly expect speed, clarity, and transparency.
That's the competitive reality. But speed without control is how lawyers get sanctioned, embarrassed, or worse.
First: stop thinking of "AI" as a search engine
This is where smart lawyers get reckless.
Ethics guidance has been blunt: many generative AI tools are not search engines that reliably "look up" law. They generate text. They can sound confident and still be wrong—or invent things that don't exist.
And courts have made examples out of lawyers who didn't respect that reality.
Two cautionary headlines every attorney should internalize:
- •Hallucinated cases = sanctions risk. Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023).
- •Fake citations—even if they came from a "helpful" AI tool—still land on the lawyer's desk. United States v. Cohen, No. 18-cr-602 (JMF), 2024 WL 1193604 (S.D.N.Y. Mar. 20, 2024).
You can be innovative without being reckless. But you can't be reckless and expect to stay credible.
The ethics framework is clearer than people think
The best way to understand AI ethics is this: Treat AI like a powerful, eager, nonlawyer assistant with a history of occasionally making things up.
That means: use it, but supervise it, verify it, and never outsource judgment to it.
Across ABA and multiple jurisdictions, the recurring obligations look like this:
Competence
Know what the tool is doing, and what it can't do
Confidentiality
Protect client information; understand data retention and training risks
Communication / Consent
Some contexts call for client disclosure or informed consent
Candor and truthfulness
You own what you file; 'AI did it' is not a defense
Supervision
Policies, training, and oversight—firmwide
Fees
Don't bill for time you didn't spend; don't turn 'AI efficiency' into hourly inflation
That's the backbone. Now let's get practical.
What every attorney should do now: the 10-point AI playbook
Create an "AI Allowed Uses" policy (one page beats no page)
If you don't define the rules, your team will define them for you—by accident. Include:
- • Approved tools (and explicitly banned tools for client data)
- • What may/may not be entered into prompts
- • Who can use AI for what tasks (attorney vs. staff)
- • Required verification steps before anything leaves the firm
Decide what "confidential" means in your AI workflow
Texas and D.C. ethics guidance highlight the risk that some tools may store inputs or incorporate them into future outputs. If you aren't "reasonably satisfied" the tool protects confidentiality, don't feed it sensitive facts without client consultation/consent.
Vet vendors like you're hiring a bookkeeper with your trust account password
At minimum, you should know:
- • Whether your prompts are used for training
- • Data retention period
- • Access controls and audit logs
- • Security posture and incident response
Separate "public AI" from "practice AI"
Not every AI tool belongs in client work. Some "consumer" tools are fine for general education or brainstorming, but not for client-specific analysis—especially where confidentiality or hallucinations create risk.
Build a verification protocol (and actually use it)
My rule: If it's going to a client, court, or opposing counsel, it gets verified like a junior associate wrote it. Verification includes:
- • Pull the cases
- • Check quotes against the source
- • Confirm statutes/regulations from authoritative databases
- • Confirm facts from the record, not the model
Treat AI output as a "drafting accelerator," not a "decision-maker"
A strong use of AI is: first-pass structure, issue lists, chronology, summarization, drafting frameworks.
A dangerous use is: "tell me the answer" and then you paste it into a filing. Ethics guidance is consistent: you're still accountable for the final work product.
Don't let AI quietly rewrite the billing relationship
The billing rule is simple: If you bill hourly, you bill the time you actually spent, not the time AI would have taken a human. This is one reason AI adoption will keep pressuring firms toward flat-fee, value-based, and hybrid models.
Decide when client disclosure is required (and document it)
Disclose (and consider consent) when:
- • You plan to input confidential information into a system with uncertain data protections
- • AI materially affects strategy, confidentiality risk, or billing/expenses
- • A client's engagement terms require it
Track court requirements like you track filing deadlines
Courts have moved in different directions: some require disclosure, some require certification, some just hammer Rule 11 and professional responsibility. Even where AI use is permitted, courts often emphasize: you remain responsible for accuracy and truthfulness.
Train your people—because "AI literacy" is now risk management
Washington's advisory opinion explicitly addresses not only generative AI, but emerging "agentic" systems that can act more like autonomous assistants. That evolution makes training and oversight even more important.
Where AI actually helps—without wrecking your license
Used correctly, AI is a multiplier in these areas (with human review):
The key is staying honest about what it is: a drafting and analysis assistant, not an authority.
Three "safe prompt" patterns that reduce risk
These are not magic. They're guardrails.
Pattern 1: "Summarize and point out gaps"
Pattern 2: "Create options, not conclusions"
Pattern 3: "Draft like a junior—then I'll review"
Notice what's missing: "find me cases" in a general-purpose model. If you want research support, use a legal-grade tool and still verify.
Final thought: AI doesn't change the profession—it exposes it
AI will make good lawyers better. It will make sloppy lawyers faster at being sloppy.
Your reputation in court is still built the old-fashioned way: accuracy, credibility, preparation, and judgment. AI can amplify those strengths, but it will also amplify weak habits.
Your clients have rights. Your practice has rules. AI doesn't change either.
Reusable prompt: Generate an AI policy for your firm
Copy this prompt to get started:
AI Disclaimer: While AI tools may have been used to assist in the preparation of this article, all content has been reviewed and verified by a human for accuracy and completeness. The views expressed are those of the author.