AI claims in your disclosure are now an enforcement surface
The SEC has been explicit that overstated AI capability claims — "AI washing" — are actionable. If AI appears in your filings, marketing, or earnings calls, treat every claim as a disclosure that must be substantiated.
The SEC's first AI-washing actions targeted investment advisers that marketed AI capabilities they did not have, and the message has carried well beyond the advisory space: statements about what your AI does are securities disclosures like any other. They must be accurate when made, substantiable on request, and consistent across every channel — registration statement, annual report, investor deck, and the company blog.
Where companies get caught
- Capability inflation. Describing a rules-based or vendor-supplied tool as "our AI" or "machine learning at scale." If the description would surprise your own engineers, it will not survive an examiner.
- Inconsistency between filings and marketing. A 20-F that says AI use is "exploratory" while the sales site promises "AI-driven outcomes" hands the staff its theory.
- Risk factors that don't match operations. Boilerplate AI risk language signals that nobody mapped actual AI use; specific, operational risk factors signal control.
- Undisclosed dependence on third-party models. If a frontier-model vendor's pricing, terms, or availability could move your numbers, that dependence belongs in risk factors and MD&A.
What to do now
- Inventory every AI claim the company makes publicly — filings, decks, website, press — and tie each to evidence: what system, what data, what measured result.
- Reconcile the registration statement or annual report against that inventory before the next filing, not after a comment letter.
- Rewrite AI risk factors from your actual architecture — model vendors, data pipelines, human review points — rather than from a peer's filing.
- Add AI statements to disclosure controls, so new claims route through legal review the same way financial metrics do.
Issuers preparing an IPO should expect AI-related comments as a matter of course; companies already reporting should assume the next periodic filing is the place to true things up. The posture is still developing, so confirm the current state of guidance before your next filing.