“Anonymize your AI inputs”

“Don’t put sensitive data into AI.”

I keep seeing this advice in our industry — from compliance consultants, industry publications, and AI itself. “Don’t put private information into AI tools.”

It might sound like good advice, but it’s actually quite restrictive. It’s the kind of reflexive take you get from disconnected parties whose primary interest isn’t helping you serve your clients better.

Where I’m not seeing this advice is from the SEC — the actual regulators. Their current guidance is more general: understand your tools and how your data is handled, stay compliant with the existing rules, and fulfill your fiduciary duty to clients. I think that’s the right approach.

There are real security, privacy, and compliance considerations around AI tools. But the “just avoid it” advice is lazy. It mirrors the early resistance to cloud storage — I was reading about how when enterprise adoption of AWS and other cloud solutions started picking up in the late 2000s, the default position was: “Don’t put sensitive data on someone else’s servers.”

Initially the caution was understandable. But the technology matured, and responsible cloud use became normal. Nobody today tells you not to save data in your cloud storage system or CRM just because you aren’t running the servers in your office closet.

It makes sense to exercise caution when sending data somewhere you don’t fully control. But the answer shouldn’t be “don’t do it.” The answer is: develop the practices that make it secure and private. And that’s where we are with AI.

What does responsible use actually look like?

• Use business/enterprise plans, which don’t retain or train on your data (or, even better, use local models, which don’t send your data anywhere)
• Train your team on security and best practices
• Be transparent with clients — explain what you’re doing, why it’s safe, and how it helps you serve them better

And it does help you serve them better. Take our discovery process as an example: we collect dozens of sensitive documents — entity and estate documents, financial statements, tax returns, and so on. We can go through them manually, which takes hours. Or we can run them through a secure, private AI tool, and have a comprehensive picture in minutes.

What do you do with the hours saved? The high-level work. The integration and analysis. The human touch. The client gets better results, and you spend your time on what really matters.

The firms treating this as a problem to solve, rather than with reflexive distrust, will pull ahead. The cloud skeptics of the late 2000s didn’t stop cloud adoption — they just slowed themselves down.

This will play out the same way.