What Is AI Washing, and How to Spot It
AI washing is marketing that overstates what a product's AI actually does, and regulators have started fining companies for it. Here are the real cases and a checklist to spot the pattern yourself.
AI washing is marketing that overstates what a product's artificial intelligence actually does. It shows up as claiming AI where a human is doing the work, claiming a sophisticated model where there's a simple script, or citing benchmark results nobody can reproduce. The term echoes "greenwashing," and regulators have adopted it directly. Both the U.S. Securities and Exchange Commission and the Federal Trade Commission have brought enforcement actions describing exactly this conduct since 2024. This is not a theoretical risk. Below are the real cases, plus a short checklist for spotting inflated AI claims before you buy, invest, or write about a product.
The enforcement record so far
On March 18, 2024, the SEC charged two investment advisers, Delphia (USA) Inc. and Global Predictions Inc., in what it called its first AI washing enforcement actions. Delphia had told the public its "proprietary algorithms" combined client data with market data to predict which companies were about to take off. The SEC found Delphia never built that capability and never used client data for it. Global Predictions marketed itself as the "first regulated AI financial adviser" and said its platform generated investment recommendations, but the underlying chatbot did not actually produce any. Both firms settled without admitting or denying the findings, paying $225,000 and $175,000 in penalties respectively.
Six months later, on September 25, 2024, the FTC opened "Operation AI Comply," a sweep of five cases against companies it said were using AI hype to deceive customers. One target, DoNotPay, had marketed a chatbot as a substitute for a human lawyer. The FTC's complaint said the company never tested whether the bot's output matched what a real lawyer would produce and never employed or retained any attorneys to check its work. DoNotPay settled without admitting wrongdoing.
The pattern surfaced again in April 2025, when the SEC and the U.S. Attorney's Office for the Southern District of New York charged Albert Saniger, founder and former CEO of shopping app Nate Inc., with securities fraud. Saniger had told investors that Nate's app used AI to complete online purchases automatically, without human involvement, and raised more than $42 million on that claim. According to the SEC's complaint, the checkout process was actually carried out by contract workers overseas who manually entered order details. Saniger has denied the allegations and the case is pending in court.
Why the gap exists
None of this requires the underlying product to be useless. Nate's app worked; humans did the work that got marketed as AI. Delphia had real client data on file; it just never built the algorithm it advertised. The gap regulators keep finding sits between what a pitch deck says a system does on its own and what a person is actually doing behind the interface. That gap matters past the fundraising stage too. When a customer can't tell whether a decision came from a model or a human, figuring out who is responsible when the system gets something wrong gets a lot harder.
A checklist for spotting inflated AI claims
Vague "powered by AI" with no specifics. If a product page or pitch can't name the task the AI performs, the input it uses, or the decision it makes, the phrase is doing marketing work, not descriptive work.
No named model or vendor. Legitimate products will say whether they're built on a specific commercial model, an open-source model, or in-house tooling. Silence on this exact point is often deliberate. How to vet an AI vendor covers what documentation a buyer should be able to ask for directly.
Benchmark claims with no methodology. A number like "95% accuracy" or "10x faster than a human" means nothing without the test set, the comparison baseline, and who ran the test. See what actually counts as an AI benchmark before taking a vendor's chart at face value.
Human-in-the-loop work marketed as full autonomy. Every case above turned on this exact question. Ask directly what share of the output a person reviews, edits, or produces before a customer ever sees it.
No disclosed failure mode. Real AI systems make mistakes. A vendor who claims theirs doesn't, or won't describe what happens when it's wrong, is selling a marketing claim rather than a product. The same instinct that catches this also catches outright AI scams, just at smaller scale.
What to actually do with it
Ask for a live, unscripted demo rather than a produced video.
Ask in writing which model or vendor powers the feature, and get a specific name, not a category.
Ask what percentage of the output requires human review before it ships to a customer.
If the company is public, check its SEC filings and risk disclosures for how it describes its own AI use, not just its press releases.
Treat an AI claim like any other unverified vendor claim inside the broader set of AI risks: ask for evidence, not confidence.
None of this is an argument against using AI. It's an argument for precision. A product that uses a real model for a real task, disclosed plainly, doesn't need embellishment. Every case above involved a company that already had a working product and paying customers, and chose to inflate the claim anyway. That's the tell: overstatement is rarely about covering a total absence of technology. It's about covering the gap between "we built this" and "we bought an API key and hired some contractors."
FAQ
What is AI washing in simple terms?
AI washing is when a company markets a product as using artificial intelligence more extensively, autonomously, or effectively than it actually does, similar to how greenwashing overstates a company's environmental practices.
Is AI washing illegal?
It can be, when the claims are made to investors or consumers. The SEC has charged AI washing conduct under securities fraud law, and the FTC has used its authority over unfair or deceptive practices under Section 5 of the FTC Act. Both target false or misleading claims, not the use of AI itself.
How can I tell if a company is faking its AI claims?
Look for vague powered-by-AI language with no named model or vendor, benchmark numbers with no stated methodology, and marketing that describes full autonomy where a human is plausibly doing review or data entry behind the scenes.
Has the FTC or SEC actually punished companies for AI washing?
Yes. The SEC fined Delphia and Global Predictions in March 2024, the FTC brought five cases under Operation AI Comply in September 2024, and the SEC and DOJ charged the former CEO of Nate Inc. with securities fraud in April 2025 over AI-related claims to investors.
What's the difference between AI washing and normal marketing hype?
Ordinary hype exaggerates value. AI washing misrepresents specific, checkable facts, like which model powers a product, whether a human completes the task, or whether a benchmark result is real, that a regulator or buyer can later verify as true or false.
How did this land?
About the author
Senior Editor, AI & Product
Cecilia leads the Swarmz editorial desk. She has spent a decade turning complex AI and product topics into writing people actually finish, and she owns the blog's quality bar.
