Nearly every AI startup says it uses AI. Whether that is true, and in which step of the product, is a factual question, and the US Securities and Exchange Commission has now brought cases where the answer was allegedly or admittedly "not as described". This article reads four SEC releases to see what AI washing looked like in each, and turns them into questions a customer, investor or journalist can ask. It is analysis of public releases, not legal advice, and it is not a claim that any company named here is typical.
What the SEC means by AI washing
The SEC's June 2024 release on Joonko quotes the Director of Enforcement saying the agency "will continue to police the markets against AI-washing" and urging investors to "beware of companies exploiting the fanfare around artificial intelligence."[2] In our reading, the term covers a spread of conduct: claiming AI that does not exist, overstating what real AI does, and using AI vocabulary to sell a conventional fraud.
Case 1 and 2: two investment advisers (March 2024)
On 18 March 2024 the SEC announced settled charges against Delphia (USA) Inc. and Global Predictions Inc., both investment advisers rather than startups, with civil penalties of $225,000 and $175,000. Both neither admitted nor denied the findings, and both were censured.[1]
- Delphia: from 2019 to 2023 it made claims in SEC filings, a press release and on its website that it used AI and machine learning incorporating client data. The SEC found it lacked those capabilities.[1]
- Global Predictions: in 2023 its website and social media called it the "first regulated AI financial advisor" and claimed "expert AI-driven forecasts." The SEC also found a Marketing Rule violation over a claim about tax-loss harvesting.[1]
The SEC's Chair at the time said: "Investment advisers should not mislead the public by saying they are using an AI model when they are not."[1]
Case 3: Joonko (June 2024, allegations)
The SEC charged the founder and CEO of Joonko, an AI recruitment startup, alleging at least $21 million was raised through false statements. The release says the founder claimed more than 100 customers including Fortune 500 companies, revenue above $1 million, and more than 100,000 active candidates, and later supplied falsified bank statements and forged contracts to a suspicious investor. The SEC sought injunctions, penalties, disgorgement and an officer-and-director bar; criminal charges were announced in parallel.[2]
The Enforcement Director called it "an old school fraud using new school buzzwords." In our reading, that matters: the alleged lies were about customers, revenue and users, which any diligence process checks regardless of AI. These are allegations, not findings, and we did not read the outcome.[2]
Case 4: Nate (April 2025, allegations)
The SEC's litigation release says that between spring 2019 and December 2022 the founder of Nate, Inc. marketed a shopping app as using AI to process transactions, and told Seed and Series A investors that purchases were completed without human involvement. The SEC alleges the app relied heavily on contract employees who manually entered orders, had a lower transaction success rate than stated, and could not use AI to complete purchases. It alleges more than $42 million was raised and charges violations of Section 17(a) of the Securities Act and Section 10(b) and Rule 10b-5 of the Exchange Act.[3] The release does not mention a parallel criminal case. The SEC's fiscal-year 2025 results release lists the Nate matter among its notable cases.[4]
What the four cases have in common
| Question | Delphia / Global Predictions | Joonko | Nate |
|---|---|---|---|
| What was claimed | Use of AI; AI on client data; "first regulated AI financial advisor" | Customers, revenue, candidate numbers | Purchases completed by AI without humans |
| Status | Settled findings | Allegations | Allegations |
| Checkable by | Asking for the model, data flow and capability evidence | Customer references, bank and contract evidence | Observing orders and auditing volumes against staffing |
A practical checklist (our analysis)
These questions are our synthesis, not SEC guidance.
- Where exactly is the model in the workflow? Ask for a diagram of one request from input to output, with each step labelled as model, rule, or person.
- Which data does it use? The Delphia finding concerned a claim about client data. Ask what data is read, when, and where it is stored.
- Who does the work on a bad day? The Nate allegation turns on humans handling orders. Ask what share of transactions needs human handling, and how that is measured.
- Can you watch it, not just see a demo? Ask to run your own cases, or to inspect logs for a sample of real transactions.
- Do the ordinary numbers reconcile? For Joonko the alleged problems were customers and revenue. Ask for contracts and bank-confirmed receipts under normal diligence.
- Is the regulated claim wording defensible? Terms like "first", "regulated" or "AI-driven" are factual claims. Ask what evidence supports each.
What remains uncertain
We did not read the underlying orders or complaints in full, only the SEC releases, and we did not check later outcomes for Joonko or Nate. The FY2025 results release also states the current Commission's view that enforcement should not be used for policymaking, so how aggressively AI-washing theories beyond plain fraud will be pursued is not something these documents settle.[4] For checking a raise, see the Form D guide, and for reading performance claims, how to read an AI benchmark claim. More coverage lives on the startups topic page.
