The moment you paste a client's facts into a chatbot to "just draft something quickly," you may have shared confidential information with a third party — and possibly with the model that trains on it. Confidentiality is the oldest duty in the profession, and generative AI puts new pressure on it. Using AI to draft is fine; being careless with what you feed it is not. Here is how to keep client confidences safe.
The duty hasn't changed — the surface area has
Under Model Rule 1.6, a lawyer must not reveal information relating to the representation without informed consent, and must take reasonable steps to prevent inadvertent disclosure. ABA Formal Opinion 512 (July 2024) applies this squarely to generative AI: before entering client information into a tool, a lawyer must consider where that information goes, who can access it, and whether it will be used to train the model. The concern is concrete — the ABA has warned directly about accidental disclosure of client and privileged information through AI tools.
Why consumer chatbots are the risky default
Free, general-purpose chatbots are the tools most lawyers reach for first — and often the worst fit for confidential work. By default, many consumer AI services may retain your inputs and use them to improve the model. That means client facts you paste in could persist on a third party's servers and, in the worst case, surface elsewhere. A convenient tool is not automatically a confidential one.
Practical safeguards
- Minimize and anonymize. Enter only what the task needs. Strip names, identifiers and specifics where a generic draft would do just as well.
- Choose tools with the right data terms. Look for training-data opt-outs, clear data-retention limits, and business/enterprise terms — not consumer defaults. Evaluate the vendor's handling before you trust it with client facts.
- Get informed consent when appropriate. If a task genuinely requires sharing sensitive client information with an AI tool, consider whether Rule 1.6 calls for the client's informed consent.
- Write it into your firm policy. Decide in advance which tools are approved and what may be entered — see the playbook in AI document drafting for small law firms.
- Separate confidentiality from accuracy. Protecting data is one duty; verifying output is another — don't skip the review checklist.
What "responsible by design" looks like
The safest posture is a tool built for legal work, where drafting, review and signing happen in one controlled environment rather than by copy-pasting client data into a public chatbot. That is the intent behind eObig's AI assistant: help draft and edit a document you keep, review and sign — so client material stays inside your workflow. The broader ethics frame is in can lawyers use AI to draft legal documents.
In brief
- Rule 1.6 confidentiality applies fully to AI; ABA Formal Opinion 512 requires you to know where client data goes before you enter it.
- Consumer chatbots often retain and train on inputs by default — a convenient tool is not automatically a confidential one.
- Minimize what you input, choose tools with proper data terms, get informed consent when needed, and codify it in a firm policy.
- Confidentiality and accuracy are separate duties — do both.
General information, not legal advice — follow your jurisdiction's rules. See drafting that keeps client material in your workflow at eObig's AI features or book a demo.