In March 2024, former SEC Chair Gary Gensler released a video in which he officially coined the term AI washing. The definition was precise: the practice by which companies exaggerate or misrepresent their artificial intelligence capabilities to attract investors, customers or talent.
The term was new. The phenomenon was not.
As early as 2019, a study by MMC Ventures analysed 2,830 European startups presenting themselves as “AI startups”: 40% did not use real artificial intelligence, but simple automation tools or traditional software rebranded under a new name. What has changed from 2022 onwards is the scale. With the public explosion of large language models, calling yourself “AI-powered” has become the most widespread and least verified claim in global technology marketing.
How it works
AI washing takes three main forms, which often overlap.
The first is outright fabrication: a company claims to use artificial intelligence in processes that are in fact handled entirely by humans. The most emblematic case is Nate Inc., an e-commerce startup that had raised over $40 million by promising investors a “fully automated, AI-based” app. In reality, hundreds of workers in the Philippines were processing every transaction manually. According to court documents, the actual level of automation was essentially zero.
The second form is systematic exaggeration: the company does use some AI tools, but vastly overstates their capabilities in its public communications. Delphia, an investment firm, promoted itself as a pioneer of AI applied to finance. In reality, its system consisted of basic data analysis, without the advanced predictive models described in its marketing materials.
The third is the semantic rebrand: products that already existed as automation tools or traditional algorithms are renamed “AI-powered” without any substantial change to their code or operation. Researchers describe this as adding “go-faster stripes” to products: superficial enhancements that capitalise on the hype without any real technological advance.
Who is doing it
The phenomenon cuts across sectors and company sizes.
In fintech and investment, documented cases are the most numerous. In addition to Delphia and Global Predictions, the SEC has brought proceedings against Rimar Capital for false statements about the AI capabilities used in portfolio management. The common pattern: promises of superior returns thanks to predictive systems that did not exist in the form described.
In e-commerce and digital services, Amazon found itself at the centre of a controversy in 2024 when it emerged that the “Just Walk Out” technology installed in Amazon Fresh and Amazon Go stores actually relied on more than 1,000 workers in India manually reviewing transaction videos.
In marketing and communications, researchers in the International Journal of Market Research have documented how competitive pressure is pushing agencies to claim AI capabilities in creative and strategic processes that in practice amount to basic use of generative tools. Tools that teams use on a standard subscription are presented as proprietary infrastructure.
The legal consequences
The regulatory framework has moved with unusual speed compared with previous technological shifts.
In the United States, the SEC built an enforcement framework in under two years. Its first actions, in March 2024, against Delphia and Global Predictions set the precedent: false statements about AI capabilities constitute a violation of the Securities Act and the Investment Advisers Act, with civil penalties, cease-and-desist orders and formal censures. The two firms paid civil penalties of $225,000 and $175,000 respectively.
In April 2025, the SEC and the Department of Justice charged the founder and CEO of Nate Inc with investor fraud. The following May, the FBI and the US Attorney’s Office for the Southern District of New York added criminal charges for attempted fraud of $40 million.
In parallel, the FTC has brought proceedings based on the prohibition of deceptive commercial practices. In the Air AI case, the FTC alleged that the voice-AI platform had made false or unsubstantiated claims about the earnings buyers would achieve, applying Section 5 of the FTC Act.
In Europe, the regulatory framework is consolidating around the AI Act. From August 2026, its transparency obligations require all AI systems to be classified by risk level. Breaches can incur fines of up to 7% of global annual turnover. The United Kingdom, under the Digital Markets, Competition and Consumers Act 2024, provides for fines of up to 10% of global turnover for misleading commercial practices linked to technology claims.
Why it happens
Startups that mention AI in their pitch materials attract between 15 and 50% more investment than those that do not. In a market where the AI signal acts as a valuation amplifier, the temptation to use it even without substance is strong.
The problem is that the signal degrades. AI washing creates expectations that go unmet, and unmet expectations produce what researchers call “AI booing” (a public rejection of AI as a category). That rejection also harms those who genuinely use AI. The erosion of trust in the claim is a collective harm, spread across the entire market.
The mechanism is identical to that of greenwashing: the term loses its signalling power, consumers and investors stop distinguishing between those who do and those who merely claim, and those who genuinely do pay the cost of the opacity created by those who only claim.
The point
AI washing is a governance problem before it is a compliance one.
A company that claims AI capabilities is, in fact, building a promise its product cannot keep, towards customers who sooner or later will experience it first-hand, and towards investors who sooner or later will demand an account of the results.
The gap between narrative and operational reality is always temporary. Regulators measure it. Customers experience it. The market prices it.
Those who govern a brand seriously know that every public statement about its capabilities is a commitment that must be demonstrable. This applies to sustainability. It applies to values. It applies to artificial intelligence.
The most expensive claim a company can make is one its own product cannot support.
New Connections (FAQ)
How do I verify whether an AI supplier really does what it claims?
Three concrete questions to ask before signing any contract. First: can I see a live demonstration of the system working on my specific use case? Second: which performance metrics are measured, and how often are they shared? Third: is there technical documentation describing the system’s architecture, including third-party components? A supplier who hesitates on any of these three questions is probably hiding something.
If we have used the “AI-powered” claim loosely in the past, what are we exposed to?
It depends on the jurisdiction and the type of statement. In the United States, statements to investors are the most exposed: the SEC is explicit in classifying AI washing as a violation of the Securities Act. In Europe, the AI Act and the greenwashing directives (the regulatory mechanism is similar) carry penalties of up to 7% of turnover. The first step is an audit of existing communications: identify the claims that are not borne out by operational reality and correct them proactively before external scrutiny arrives.
Is AI washing illegal or merely a reputational risk?
Both, in proportions that vary with the context. For listed companies and those raising capital from investors, false statements about AI capabilities can already lead to criminal prosecution in the United States, as the Nate Inc. and Presto Automation cases show. For unlisted companies communicating with consumers, the risk is predominantly regulatory (FTC, AGCM, AI Act) and reputational. In either case, the cost of a proactive correction is always lower than the cost of proceedings.
Do SMEs also face consequences for AI washing?
Yes, although enforcement currently focuses mainly on listed companies and those raising investment. SMEs that use the “AI” claim in their communications to customers are exposed to European consumer directives and the AI Act. The most immediate risk for an SME is not regulatory but contractual: if a client chose a supplier on the basis of declared AI capabilities that were not borne out, it has grounds to challenge the contract or claim damages.