Can Lawyers Use ChatGPT?

The short answer: it depends on the tier

ChatGPT is not one product with one privacy posture — it is several products with materially different data-handling commitments. The free plan, the paid Plus plan, the Team and Enterprise plans, and direct API access each have distinct rules about whether your inputs are used to train OpenAI’s models.

That distinction matters for lawyers specifically because Model Rule 1.6 requires reasonable measures to prevent unauthorized disclosure of client information. A tool that trains on your inputs by default is not a tool with reasonable safeguards for client data.

What ABA Formal Opinion 512 actually says

In July 2024, the ABA Standing Committee on Ethics and Professional Responsibility issued Formal Opinion 512, its first comprehensive guidance on lawyers’ use of generative AI. The opinion does not ban ChatGPT or any other AI tool. Instead, it maps existing ethics rules onto AI use.

The key obligations it identifies:

• Competence (Rule 1.1) — You must understand what the tool does and what it cannot do. Comment 8 to Rule 1.1 has long required that lawyers keep up with relevant technology. Opinion 512 applies that standard directly to AI.
• Confidentiality (Rule 1.6) — Lawyers must not input client information into a tool that could expose it without informed client consent or proper contractual controls.
• Supervision (Rules 5.1 & 5.3) — AI output is not self-certifying. You remain responsible for verifying everything the tool produces before filing or advising based on it.
• Candor (Rules 3.1 & 3.3) — You must verify citations and legal statements before submitting them to a tribunal. AI hallucinations are a documented and recurring problem.
• Reasonable fees — You cannot bill a client for the general time you spend learning a tool you intend to use across matters.

Formal Opinion 512 is not binding law — state bars and courts are the binding authorities — but it is treated as persuasive guidance by most jurisdictions, and it reflects the direction state ethics opinions have been moving since 2023.

The Free and Plus tier problem

ChatGPT Free and Plus default to using your conversation history to improve OpenAI’s models. That means client facts, opposing party names, case strategy, and privileged communications you type into the chat window could become training data.

You can disable this: Settings → Data Controls → Improve the model for everyone → Off. But this is a manual step that is not covered by any contractual commitment from OpenAI and provides no Data Processing Agreement. For an individual practitioner with a small number of clients, the practical risk may feel abstract — but it is real, and your bar may not see it as a “reasonable measure” under Rule 1.6.

Enterprise and API tiers: a materially different picture

ChatGPT Team, Enterprise, and the API platform operate differently. By default, OpenAI does not train on inputs or outputs from these tiers. Enterprise accounts also receive a Data Processing Agreement (DPA) — a contractual commitment that is absent from consumer plans.

This does not mean Enterprise is without risk. You still need to evaluate:

• Whether your state bar has issued specific AI guidance that imposes requirements beyond the ABA opinion
• Whether the task itself — even with Enterprise controls — is one where inputting privileged content is necessary at all
• Whether your engagement letter addresses AI tool use (several state bars are beginning to require or recommend disclosure)

A practical checklist for solo and small-firm attorneys

This is not legal advice — consult your state bar’s ethics counsel for jurisdiction-specific guidance. It is a technical security checklist based on how these tools actually handle data.

The verification requirement is non-negotiable

Regardless of which tier you use, Opinion 512 is clear that AI output requires independent professional review. ChatGPT can produce plausible-sounding case citations for cases that do not exist. It can summarize statutes inaccurately. It can miss recent case law.

If you are using AI-generated content in anything that goes to a client or a court, verify every factual and legal claim independently. This is not optional under the duty of competence or the duty of candor. The sanctions record since Mata v. Avianca demonstrates what happens when attorneys skip this step.

Frequently asked questions