SEC AI washing: questions to ask about AI use claims
Last reviewed May 24, 2026
SEC AI washing sources are useful when a company says its investment process, forecast, or advisory product uses AI in a specific way. This guide turns the source examples into disclosure questions and evidence requests.
Fastest path: copy one exact vendor sentence that matches this pattern, then open the checker. Add the public URL only if you want readable page context recorded alongside the wording. The result is an evidence-burden note you can reuse in vendor follow-up or internal review, not a verdict. Not sure what a result looks like? See a sample receipt.
What to verify before you rely on the claim
- A dated description of the AI capability as it existed when the public claim was made.
- A plain-language explanation of which step uses AI and which step does not.
- Comparison evidence for first, only, best, or expert claims.
Sources behind SEC AI washing
- · March 18, 2024
Source for Delphia and Global Predictions AI use statements.
- · March 18, 2024
Source for Delphia AI and machine-learning representations about client data, predictive investing, and advertising controls.
- · March 18, 2024
Source for Global Predictions first-regulated-AI-advisor, AI-driven forecast, performance, and marketing-rule claim context.
Documented SEC AI washing examples
"make our artificial intelligence smarter"
Accuracy / Performance- Source and date
- SEC Delphia administrative order · March 18, 2024
- Evidence signal
- AI use claim tied to client data and predictive investment process.
- Evidence gap
- A reader needs the model role, data used, investment decision boundary, controls, and whether the described capability existed at the claim date.
- Buyer question
- For the AI smarter claim, what client data is used and what investment decision does the AI actually influence?
"first regulated AI financial advisor"
First / Only / Best- Source and date
- SEC Global Predictions administrative order · March 18, 2024
- Evidence signal
- First-of-kind claim in a regulated product context.
- Evidence gap
- A reader needs the comparison set, jurisdiction, date, registration basis, and evidence that the first claim was current when made.
- Buyer question
- For the first regulated AI financial advisor claim, what market definition and date support first?
"Expert AI-driven forecasts"
Accuracy / Performance- Source and date
- SEC Global Predictions administrative order · March 18, 2024
- Evidence signal
- Forecasting claim with expert framing.
- Evidence gap
- A reader needs forecast methodology, benchmark comparison, time horizon, error rate, and human review boundary.
- Buyer question
- For the Expert AI-driven forecasts claim, what benchmark and forecast horizon support the wording?
"use your personal data and the data of thousands of other investors to make predictions"
Vague AI-powered- Source and date
- SEC Delphia administrative order · March 18, 2024
- Evidence signal
- Client-data AI use claim without disclosing that the data collection and AI investment link described to clients was not operational at the time.
- Evidence gap
- The SEC found Delphia represented that it collected and analyzed client spending data to train its AI and generate investment insights, but the described capability did not exist as claimed. A reader needs confirmation that the AI capability described was operational when the claim was public, what client data was actually used in the model, and which investment steps the AI influenced versus which remained fully human-managed.
- Buyer question
- For the client-data AI investment claim, was the described AI capability operational when the claim was published, what client data was actually fed into the model, and which investment decisions did the AI directly influence?
Evidence map for SEC AI washing
| Claim pattern | Evidence needed | Buyer question |
|---|---|---|
| AI uses client data to improve investment decisions | Model role, data fields, governance controls, decision boundary, and update history. | Which investment step is influenced by AI, and which step remains human-controlled? |
| First regulated AI advisor or first AI product in a category | Market definition, jurisdiction, comparison set, date of comparison, and update process. | What exact category does first apply to, and when was that comparison last checked? |
| AI-driven forecasts or expert predictions | Forecast target, benchmark, backtest period, error distribution, and limitation disclosures. | How did the AI forecast perform against a non-AI baseline over the same period? |
| AI capability claim in a filing, website, email, or social post | Dated public copy, model capability at that date, data actually used, disclosure updates, and advertising review controls. | What evidence shows the AI capability existed when the public claim was live? |
Evidence buyers need for SEC AI washing
- A dated description of the AI capability as it existed when the public claim was made.
- A plain-language explanation of which step uses AI and which step does not.
- Comparison evidence for first, only, best, or expert claims.
- Forecast metrics that include baseline, period, and error distribution.
Buyer questions for SEC AI washing
- For this SEC-style AI use claim, what model, data, or workflow step is actually involved?
- Does the public copy distinguish between planned AI use and current AI use?
- If the claim says first or regulated, what comparison universe and date make that statement supportable?
- If the claim says AI-driven forecast, what benchmark shows the forecast adds value beyond a simpler baseline?
Safer wording for SEC AI washing
- Uses model-assisted analysis in specified workflow steps, with human review before portfolio changes.
- Currently uses a named analytical model for a defined workflow step; client data sources and limitations are described in current disclosures.
- Offers AI-assisted forecasting features; forecast performance varies by market conditions and time horizon.
- Registered investment adviser offering AI-assisted planning tools, without claiming first-of-kind status.