AI washing and the SEC: how PE firms should describe portfolio AI and data initiatives

The short answer on AI washing

AI washing is the practice of overstating how much a firm, a fund or a portfolio company actually uses artificial intelligence. Whether a given AI statement is a problem for a private equity adviser depends on three things: what was claimed, who it was said to, and whether the firm's own files can prove it. A workable test is to hold an AI claim to the same standard as any other statement to investors: accurate, not misleading by omission, and backed by evidence the firm can produce on request.

The SEC has publicly announced enforcement actions against investment advisers over misleading statements about their use of AI. This page does not summarize individual orders, because press summaries tend to compress the facts that mattered. Ask compliance counsel to pull the current orders and staff statements from sec.gov.

For a PE firm the exposure sits in two layers:

• The firm's own claims: AI in sourcing, diligence, underwriting or portfolio monitoring, as described in pitch books, the website, DDQ answers and conference panels.
• Claims about portfolio companies: AI-driven margin gains, AI features in products, or new data revenue, as described in case studies, quarterly letters and the value creation story told during a raise.

Both layers come under pressure during fundraising, when LPs ask how the operating model really works. The guide to how LPs evaluate a PE firm's operating partner model covers what they probe.

What do the rules say about AI and data claims?

There is no AI exception: existing standards apply to AI statements. Three strands matter for PE firms and their portfolio companies.

Statements to investors. What an adviser tells current and prospective investors is covered by the anti-fraud provisions of the federal securities laws, and a registered adviser's marketing materials are also subject to the SEC's marketing rule. An AI claim is judged the way a track-record claim is judged. Ask counsel which provisions reach each of your documents.

Promises about customer data. Portfolio companies that license records, or use them to train their own models, must square that use with what they told customers. FTC staff wrote in January 2024 that companies' promises not to use customer data for undisclosed purposes, such as training or updating models, are enforceable, whether the promise sits in a privacy policy, terms of service or promotional material. In February 2024 FTC staff added that adopting more permissive data practices through a quiet, retroactive change to terms of service or a privacy policy could be unfair or deceptive. Both posts are staff guidance, not rules, but they show where a data story can turn into a consumer protection question.

Recommendations with a financial connection. When someone at the firm publicly recommends a service while earning from it, the FTC's Endorsement Guides FAQ says a connection the audience would not expect, and that would affect how they weigh the recommendation, should be disclosed clearly and conspicuously.

How do the rules apply in common PE situations?

Most AI washing risk sits in ordinary documents, not in headline announcements.

On the last row, the explainer on how one-time license revenue is treated in ASC 820 marks sets realistic expectations for the valuation committee.

The claim ladder for an unsigned data license

The most common slip in a PE context is describing a data opportunity as further along than it is. A data license moves through distinct stages, and each stage supports a different sentence. Nothing is binding until the company agrees price and terms and signs, so anything before signature is an exploration, not revenue.

Three facts keep the wording honest. The company keeps ownership because the data is licensed, not sold. Licenses are typically exclusive for AI training for an agreed term. And the payment is one-time, so it does not belong in recurring revenue language.

How should the firm describe a referral relationship?

With the same care as any other AI claim. If the firm or someone on the operating team registers as a SourceX partner, the economics are contingent. Partners earn 25% of the eligible platform fees SourceX actually collects from the referred company's licensing deals, capped at $100,000 per referred company, and the reward becomes payable only after the buyer pays and SourceX receives its fee. No reward is guaranteed, and an introduction, a meeting or a signed agreement on its own does not trigger payment.

Two points belong in any internal memo. The reward is a share of SourceX's fee, so it is never deducted from what the portfolio company receives. And whether it offsets management fees or needs LP disclosure is a question for the LPA and the fund CFO; the guide to fund CFO responsibilities and fee reviews walks through that review.

Disclosure and consent checklist

Run this before any AI or data statement leaves the firm.

• One named approver, usually the chief compliance officer or general counsel, signs off on new AI wording.
• Each AI claim has a substantiation file: the tool, the companies involved, the period and the measurement method.
• Pitch book, DDQ, website and quarterly letter use the same approved description.
• Every data opportunity carries its claim-ladder stage, and unsigned ones carry no revenue figure.
• Portfolio company announcements about AI training uses have been checked against privacy policies and customer contracts.
• No counterparty is named unless the signed agreement permits it.
• Public posts by team members disclose any referral connection.
• Claims are retired when a pilot ends or a tool is switched off.

Schedule the reconciliation in the portfolio operations annual calendar so it happens before each LP reporting cycle rather than in the middle of a raise.

Questions to ask your compliance counsel

1. Which of our documents count as marketing or investor communications, and which standard applies to each?
2. What evidence must we keep on file for each AI claim, and who owns that file?
3. How should we describe a portfolio company's AI product when it relies on a third-party model?
4. Can a quarterly letter mention a data opportunity before signature, and with what status language?
5. Does a referral reward connected to a portfolio company need disclosure to LPs or the LPAC under our LPA?
6. What review should a portfolio company's AI or data press release go through before it is published?

This is general information, not legal, tax or financial advice. Confirm with your own counsel, tax adviser or professional body before acting; the answer depends on your registration status, your fund documents and the facts.

Is a data licensing review an AI initiative?

No. Licensing records to AI developers is not the same as using AI, so describe it as a data licensing activity, not as an AI capability of the firm or the company. Calling it AI adoption is itself the kind of stretch this page warns about. What the activity does offer is a paper trail, because every step produces a document: a screen result, an inventory, agreed terms, a signed license.

Operating teams that run the review across a portfolio can point to the companies screened, the companies introduced and the stage each one reached. The baseline is on the who qualifies page: US companies with 50+ full-time employees at peak (contractors excluded), several years of documented operations, rights to license the data and an authorized sponsor. The playbook for private equity operating partners covers how teams run the review.

Next step

Agree the claim-ladder wording with compliance first. Then sketch which portfolio companies and advisers in your circle might fit using the network opportunity finder, and register as a partner before making the first introduction.