Client Data Security: Is Your Firm's AI Safe?
The ethics duty behind every AI tool, and the vendor questions that answer it for your firm
Client data security is an ethics duty for any law firm using AI, and you can confirm a tool meets it before you ever load a real client file. The short answer to whether client data is safe with AI is that it depends on how the tool handles your files, who can see them, and whether you keep a licensed attorney in charge of the work. Under the bar rules, an AI tool counts as nonlawyer help you bring into the firm. You stay responsible for the outcome and for the confidentiality of everything the client told you.
So client data can be safe with AI when a few conditions hold. The vendor stores and moves your data in a way that meets your confidentiality obligations. The client has consented where consent is required. Every output runs past an attorney before it leaves the firm. Miss one of those and the risk sits with you, not with the vendor.
What client data security means under the ethics rules
In July 2024 the ABA issued Formal Opinion 512, its first ethics guidance on lawyers using generative AI. Rather than banning the technology, it maps your existing duties onto it. Four rules do most of the work here.
Start with confidentiality. Rule 1.6 obligates you to protect information relating to the representation and to make reasonable efforts to prevent its disclosure. Opinion 512 reads that to mean you evaluate how an AI tool handles client information before you feed it confidential facts, and in some matters you need the client's informed consent first. Those protections cover former and prospective clients, not only active ones.
Competence comes next. Comment 8 to Rule 1.1 has said for years that a lawyer's competence includes understanding the benefits and risks of relevant technology. Applied to AI, you need to know enough about a tool to judge whether it is safe for your matters. Nobody expects you to become an engineer, only to ask the right questions and follow the answers.
Then there is supervision. Rule 5.3 makes you responsible for nonlawyer assistance, and bar opinions apply it to outside vendors, not staff alone. Handing client data to a software vendor puts you in the same position as hiring a contract paralegal. You vet the vendor and you supervise how the work gets done.
Rule 1.4 rounds it out by asking you to keep clients reasonably informed about how their matter is handled, which can include telling them when AI plays a part.
An AI tool is nonlawyer assistance under your supervision, and the confidentiality duty travels with the client file wherever it goes.
The questions to ask any AI vendor before you trust it
Most of the risk in AI comes down to what happens to your data after it leaves your screen. Before you load a single client file, get clear answers to a short list.
- •Where is my data stored, and is it encrypted while stored and while moving between systems?
- •Does the tool use my client data to train its models? A tool that learns from your inputs can surface fragments of one client's information to another user.
- •Who at the vendor can access my files, and under what controls?
- •Can I delete client data on request, and what is the retention policy when I leave?
- •Will the vendor sign an agreement that reflects my confidentiality duties?
If a vendor cannot answer these plainly, that is your answer. Opaque handling of data is the same warning sign as opaque pricing, and it deserves the same response.
Why attorney supervision is the part that protects the client
Data storage is half the picture. The other half is what the tool is allowed to decide on its own.
A safe setup keeps the AI on operations and keeps the lawyer on judgment. The tool can answer a call at 9pm, gather intake facts, organize a file, or produce a first draft of a demand letter. What it should not do is give legal advice, weigh whether a claim is worth taking, decide a filing deadline, or judge a conflict. Those calls belong to a licensed attorney, and the attorney reviews and adopts each output before it reaches a client or an adjuster.
This is the review chain, and it is worth reading about in detail in how attorney supervision of AI works. When the chain holds, an AI mistake becomes a draft you catch and correct instead of a document that already went out under your name.
Where an operations layer changes the risk picture
A point tool that copies your whole matter into a separate silo widens your exposure. Every extra place your data lives is one more place to secure and one more vendor to vet, and small firms rarely have the time to police five of them at once.
Nimbus Marlowe runs as an AI operations layer for law firms that sits in front of your existing practice-management system rather than replacing it. Intake, contract review, case preparation, demand letter drafting, and collections follow-up all run through one layer, with an attorney reviewing the output at each step. One layer means one relationship to supervise under Rule 5.3, and the finished work returns to the system you already control.
Pricing follows the same logic as the data handling. Every rate is published and flat, charged per completed unit of work. Contract review runs $65, a demand letter draft runs $170, and the CRM itself is free with no invoice. No fee is ever a percentage of a recovery, and the company never holds, transmits, or takes custody of client funds. You see what your firm will pay before you sign, which is the whole idea.
Frequently asked questions
Does using AI mean my client data trains someone else's model?
Not if the tool is built for legal work and says so in writing. The real exposure lives in consumer tools that improve themselves on whatever users type in. Ask the vendor directly whether your inputs train their models, and get the answer written into the agreement.
Do I need client consent before using AI on their matter?
Sometimes. ABA Opinion 512 ties consent to how the tool handles confidential information and how much of the client's data it touches. Many firms address it once in the engagement letter and revisit it when a matter calls for more. When in doubt, tell the client, since that also satisfies your duty to keep them informed under Rule 1.4.
Who is responsible if an AI tool leaks client data?
You are, in the eyes of the bar. Rule 5.3 keeps the confidentiality duty with the lawyer even when the work runs through an outside vendor. That is why vetting the vendor and supervising the workflow matter as much as the security of the software itself.
Does Nimbus Marlowe take a percentage of my settlements?
No. Every fee is flat and published per unit of work, and it stays the same whatever the case turns out to be worth. The company never takes a cut of a recovery and never touches client funds, which keeps a clean line between your operations vendor and your fee.
The firms that get client data security right are the ones that choose their AI carefully and keep a licensed attorney in charge of every output. To see how a supervised operations layer handles your files, start free today or book a call and we will walk through the data questions with you.
Put your firm's operations on autopilot. Reviewed by you.
Nimbus Marlowe is the AI operations layer for personal-injury firms. The CRM is free, pricing is flat and per-unit, and every output is a draft your attorneys review and adopt.