How to Use AI Without Putting Client Data in a Public Tool
Regulated firms do not have to choose between banning AI and leaking client data. Here is the third path: enterprise-grade plans with contractual controls, governed by a classification your team can actually follow.
Most leaders at regulated firms think they have two options with AI. Ban it and watch the work get slower than competitors who use it, or allow it and hope nobody pastes the wrong document into a public chatbot. Both options are bad, and neither is necessary.
There is a third path. You give your team AI on terms your firm chose and controls, instead of terms a consumer app dictates.
Why pasting into a public tool is the actual problem
When an employee drops a client document into a public AI tool, that text leaves your control. For a firm handling regulated, privileged, or non-public information, that single action can breach a client commitment, a protective order, or a regulatory obligation. The employee was not being careless. They were trying to get work done faster, and the tool made it easy.
The risk is not AI itself. The risk is the data leaving on terms nobody reviewed. Once you frame it that way, the solution gets clear: control which tools your data can touch, and on what terms.
The third path: enterprise-grade AI under real governance
The business tiers of the frontier tools are a different product from the consumer apps that share their logo. An enterprise-grade plan, Claude Teams or Claude Enterprise among them, gives the firm a commercial agreement with contractual no-training commitments, retention and access controls, admin visibility, and audit logs. Client data is still processed by the provider, but under terms your firm chose, reviewed, and can point to later, the same position you already take with hosted email.
The plan alone is not the control. Around it sits governance: a data classification that tells every employee which class of information can touch which tool, an approved-tools list with everything else out of bounds, human review before AI-assisted work reaches a client, and training so people can apply the rules in the moment. The decision is made by policy, not by hoping an employee gets it right.
What this looks like in practice
The sequence is consistent across firms. Map which of your data is actually sensitive and which is not. Put the sensitive classes behind an approved tool on enterprise terms, or keep them out of AI entirely. Approve a short list of tools, train the team, and record a certification. The person at a keyboard on a deadline never has to guess, because the policy already decided.
A small number of firms carry obligations so strict that certain data cannot be processed by any third party at all. Private deployments, models running inside a firm's own cloud tenant or on its own hardware, exist in the market for exactly that slice. They are a heavy lift to run well, and most regulated firms never need one. If you think you might be the exception, that is a deliberate decision to make with an independent advisor, not a starting point.
Where to start
The first step is mapping which of your data is actually sensitive and which is not, because that map is what every other control enforces. From there the plan choice, the approved-tools list, and the training follow in a defensible order. The right setup depends on your data, your budget, and your obligations.
Key takeaways
- The risk is not AI. It is client data leaving your control through a public tool.
- Enterprise-grade plans carry contractual no-training commitments and retention and access controls the consumer apps do not.
- A data classification and an approved-tools list decide what can touch which tool, so nobody guesses at a deadline.
- You keep frontier capability instead of banning AI and pushing people onto personal accounts.
- Private deployments exist for the rare firm whose data cannot be processed by any third party. Most firms never need one.
Talk it through
Wondering where your firm's line sits? Start with a 45-minute call.