Don't trust, verify: AI-generated compliance answers

Originally posted January 4, 2024 by Chris Vanderwolk

Generative AI is awesome - an incredibly powerful tool. Still, the prudent move is to treat the output as suspect until you independently confirm the information.

I've had people forward me AI-generated answers to compliance questions that cite to statutes. When I then went and checked the statute, it was blank.

In the cold war, the mantra was "trust but verify." With Gen AI, the mantra probably ought to be "don't trust, verify."

Originally posted on LinkedIn, where the discussion and source links live in the comments.

About the author

Chris Vanderwolk is Director of Compliance and Innovation at OneDigital | Kistler Tiffany Benefits General Agency, where he helps brokers and employers navigate the regulatory complexity of employee benefits. An ERISA attorney with more than 19 years in the benefits industry, he specializes in translating what the law actually requires into language people can use.