Shortly after the release of OpenAI’s ChatGPT in 2022, news exploded and interest in generative AI saw a massive resurgence, sparking a hype cycle that continues to peak with each new model or functionality. Suddenly, it seems like every tech company or startup is releasing an AI product that promises to revolutionize consumer behavior.
According to a study by an investment fund for new technology companies, the number of startups mentioning AI in their pitches increased from 10% in 2022 to more than 25% in 2023. More than 50% of S&P 500 companies referenced AI in their earnings calls last year, according to a report by NBC News.
But how many of these companies actually practice what they say? Last November, a U.S. Census Bureau survey found that 4.4 percent of American companies used AI to produce goods and services. A 2019 survey by London-based venture capital firm MMC found that 40% of European AI startups were not using any AI at all.
Tech companies and startups that present themselves as using AI, but do not, are the basis of “AI whitewashing.”
What is AI washing?
AI washing is a term derived from greenwashing, where companies exaggerate their environmental friendliness to please customers. Similarly, companies that claim to have integrated AI into their products, when they are actually using less sophisticated technology, can be accused of AI washing.
Advertisements that overstate the capabilities of a company’s AI tools or mislead consumers about features that are not yet operational in their AI products also constitute AI washing.
Although it is unclear who exactly coined the term AI washing, it was popularized by the United States Securities and Exchange Commission (SEC) when it imposed fines worth $225,000 ( 1.8 crore rupees) and $175,000 (1.4 crore rupees) to investment advisory firms Global Predictions and Delphia in March 2024. The securities regulator found that the companies made false representations to customers about providing “expert AI-based forecasts” and using machine learning to manage retail client portfolios.
What are concrete examples of AI washing?
The rapid advancements and vast potential of AI have led many companies, including a few tech giants, to cut corners when it comes to deploying their AI-powered products. For example, Google revealed Gemini last year with a demo video meant to show off its flagship multimodal AI chatbot, including its ability to recognize real-world images and objects. In one part of the video, a person draws a picture of an animal that Gemini correctly guesses is a duck.
However, it turned out that the video was not shot in real time. In fact, Google has confirmed Bloomberg that the video was made by sending text prompts to Gemini and stitching together still images. While the clip’s description on YouTube includes a disclaimer, there is no such disclosure in the video itself.
More recently, Amazon reportedly removed its cashierless payment systems from many grocery stores after Business Insider They found that the “Just Walk Out” technology, which claimed to use AI and sensors to detect what was in a customer’s cart, actually relied on employees in India to review transactions.
AI-related claims from multinational brands such as McDonald’s and Coca Cola have also come under scrutiny. McDonald’s recently decided to abandon its AI technology at drive-thru restaurants in the United States after customers complained that their orders were taken in error. Meanwhile, Coca Cola attempted to jump on the AI hype bandwagon last year by introducing a limited edition, AI-generated cold drink flavor, which ultimately failed to impress many customers.
There has also been a wave of AI apps that boast chatbot functionality when in reality they are just ChatGPT wrappers, meaning the underlying technology powering the apps does not belong to them but to OpenAI.
In India, Ola founder Bhavesh Agarwal’s startup released a beta version of Krutim AI that was touted as a local rival to ChatGPT. However, many users began to wonder if Krutim AI was a ChatGPT wrapper after the chatbot allegedly confirmed to them that it was “built by OpenAI.” In response to the speculation, the startup said it investigated the issue and found that the root cause was “a data leakage issue from one of the open-source datasets used in the Large Language Model (LLM) fine-tuning process.”
Even the electronics manufacturing industry has not been spared from the washout effect of AI. For example, French company Kolibree launched an AI-powered toothbrush in 2017 that essentially uses sensors to detect how users brush their teeth, after which “deep learning algorithms” analyze the data to provide personalized recommendations on how users can improve their brushing. According to reports, the annual CES 2024 consumer show held in Las Vegas, US, saw a proliferation of gadgets stuffed with an LLM and sold under the banner of generative AI.
The rush to brand itself as an AI company is part of a long-running trend of companies seeking to capitalize on the hype surrounding new and emerging technologies. At the height of the crypto boom in 2017, Long Island Iced Tea Company rebranded itself as Long Blockchain Corp only to have the U.S. SEC deregister the beverage maker’s stock in 2021 because it allegedly never went live with its blockchain business.
Why is AI washing a growing concern?
While AI washing may seem like harmless hyperbole, it could have far-reaching consequences for consumers and the tech industry itself.
Linda Yao, LenovoThe vice president of AI solutions and services, explained to ZDNet AI “washing” diverts attention and resources from leading practical AI innovation. Instead of developing meaningful AI capabilities, companies risk making ill-advised investments in superficial improvements, slowing real progress in the technology.
Yao also pointed out that AI washing can complicate decision-making for companies that are genuinely looking for valuable AI solutions. It can hamper their digital transformation efforts, stifle innovation and compromise performance, she said. When it comes to consumers, the Lenovo executive said substandard AI technology could pose data security and privacy risks, while also deterring consumers from using the technology.
To avoid AI laundering (even unintentionally), the U.S. Federal Trade Commission (FTC) recommends that companies self-monitor by asking themselves key questions such as: “Are you exaggerating what your AI product can do? Are you promising that your AI product does something better than a non-AI product? Does the product actually use AI?”
“Before you label your product as AI-powered, also note that simply using an AI tool in the development process is not the same as a product containing AI », added the FTC.
At the same time, the Securities and Exchange Board of India (SEBI) had somewhat warned against AI washing in a 2019 circular.
“As most AI/ML systems are black boxes and their behavior cannot be easily quantified, it is imperative to ensure that any financial benefits advertised through these technologies in financial products intended for investors and offered by intermediaries do not constitute a misrepresentation,” the circular states.
