If AI touches advice or decisions, you have to be able to explain them.
AI drafts client emails, summarises meetings, screens applications and suggests investments. None of that changes your obligations. Regulators have said so plainly, and the first enforcement actions have already landed on small firms.
The rules and the numbers.
The rules haven't changed for AI.
FINRA reminded member firms in June 2024 that its rules are "technology neutral": supervision and communications rules apply to AI just as they do to anything else.
Source: FINRA Regulatory Notice 24-09
Saying you use AI is a claim you have to back up.
The SEC has fined advisers for describing AI they didn't have.
See the case below.
States are testing the AI that lenders use for bias.
Massachusetts settled with a lender over biased AI underwriting in July 2025.
See the case below.
Real cases.
Earnest Operations, 2025
Massachusetts' Attorney General found that a student lender's AI underwriting tools penalised Black and Hispanic applicants, partly by using each applicant's college default rate, and automatically rejected applicants based on immigration status. The lender hadn't tested its AI for this. It paid $2.5 million and must now keep written AI policies and report to the Attorney General.
If you can't show you tested it, you can't defend it.
Delphia and Global Predictions, 2024
The SEC fined two investment advisers a combined $400,000 for "AI washing". One said it used AI to analyse clients' data to pick investments, and hadn't built it. The other called itself the "first regulated AI financial adviser" and couldn't back it up.
Every claim about AI in your marketing needs evidence behind it.
Where it goes wrong, and how we fix it.
Figures and advice that no person checked.
AI is fluent and confident, including when it's wrong.
Plus puts a review step in your procedures for anything AI drafts before a client sees it, and trains your team on what to check: figures, sources, names and tone.
Client data in AI tools without a data agreement.
Pasting a client's details into a free chatbot can mean they're stored or used for training.
Plus lists the tools approved for client data and what can go into each, and allows client data only where there's a data agreement. Starter flags tools on your website whose terms let them train AI on your data.
Decisions nobody can explain.
If AI helps decide who gets credit, at what price, or which product they're offered, someone has to be able to explain why.
Plus requires a person who reviews and can explain any decision AI helps make, and says what to record so you can show it later.
AI claims in your marketing.
"AI-powered", "smart" and "personalised" are claims.
Starter flags claims about AI on your website that you may not be able to back up, and suggests plainer wording.
What it looks at.
Your website's text and claims, the tools it loads and what they do with data, your privacy notice, and how you say you use AI. Score, top three gaps and a plan recommendation in about 30 seconds.
See what your website says about how you use AI.
Your website, a few questions, and the three things to fix first.