AI is on everyone’s lips as the excitement for AI’s potential continues. Players are not only talking about AI but actively engaging in developing, integrating, selling and marketing it. While companies have developed impressive AI models which are shaping the future right now, regulators and law enforcement are raising suspicions that like any new ‘hot topic’, AI has created a gold rush atmosphere allegedly attracting false claims. The accusation reads: Not everyone who wants part of the action might actually be what they seem.
Regulators are looking into the assumption that some sellers put a little too much gloss on their products by exaggerating their capabilities in an attempt to keep up with the race for market share and shareholder interest. At the same time, this harms all those who are actually leading AI development. The situation results in regulators starting to sense false claims anywhere AI is marketed. After greenwashing authorities have now found a new playground for investigations: so-called AI washing.
A familiar phenomenon: Parallels with alleged Greenwashing
We have seen this type of phenomenon before with the rise of sustainability and green products. Initially, there was little understanding and no definition of what was to be expected of a ‘sustainable’ product. But as soon as that understanding grew authorities started investigating companies for alleged greenwashing. Today greenwashing is on the radar of law enforcement agencies around the world. The same must be expected for behaviour labelled AI washing. As public interest and understanding of AI grows, so does the risk of high scrutiny, public investigations and even criminal liability for AI washing. This development can bring risks to any company marketing and selling AI – regardless of how well-intentioned they are.
Risks of AI washing investigations: Suspicion is sufficient
Cases of actual AI washing can i.a. constitute fraud, for example in cases in which customers do not get a product providing the AI capabilities advertised and therefore may be worth less than they paid for. Or investors may have only invested in a company because it claimed to be ahead of the curve in AI development, when in fact it is not. Since the nature of any investigation is to start with a mere suspicion that needs to be cleared up, investigations will inevitably affect companies not engaging in any AI washing but are wrongly accused. Another factor that increases the risk of investigations is that there is still no agreed definition of AI yet. This can lead to investigations simply because there is a lack of understanding of what AI is or is not. Even if an investigation does not lead to any charges, it still can damage a company's public reputation. It also involves the risks to compromise trade secrets such as the AI model itself.
The two-fold role of the European AI Act
While the risks for investigations increase with growing public interest, there are also mitigating developments. The European Union passed the AI Act in March 2024. It should provide for a broader understanding of AI and thus limit the risks of investigations due to mere misunderstandings. Although the act’s definition of AI only determines the types of applications to be regulated by the AI Act, a consistent understanding of AI in this respect may have a spill-over effect in other areas.
However: While an emerging understanding of what AI is should reduce confusion, it in itself carries a risk of more regulatory scrutiny. A growing public understanding provides more reason for alleging fraud. For example, considering a case of AI washing, where a product is claimed to be 'powered by AI': If purchasers of such a product have no expectations associated with ‘AI’ they cannot be misled. Therefore, as purchasers’ expectations grow more specific, so does the risk of investigation for alleged fraud. Further, the AI Act establishes additional authorities to oversee that providers of AI systems comply with the AI Act. Thus, a dedicated authority may also look for alleged AI washing in the form of low-quality AI claimed as high-performance AI.
The SEC in the US: AI washing investigations move into action
The US have already moved in this direction. In March 2024 the US Securities and Exchange Commission (SEC) announced settled charges against two investment advisers, for allegedly making false and misleading statements about their use of AI. The companies had advertised that they used AI in their investment processes, for example to analyse client data and then to predict which companies and trends ‘are about to make it big’. Following its investigations, the SEC concluded that neither firm had the AI capabilities they claimed. In each case, the charges were civil penalties, to which the companies agreed with the SEC without admitting the SEC’s findings, and are relatively small compared to other charges – including civil penalties – that ran into the millions and billions of dollars.
Still, the understanding of AI and AI washing is only starting. More allegations and criminal charges should be expected. A warning statement by SEC’s Enforcement Director, Gurbir Grewal in relation to the settlements set the tone accordingly: ‘Today’s enforcement actions should serve notice to the investment industry, that if you claim to use AI in your investment processes you must ensure that your representations aren’t false, they aren’t misleading.’
Preparation is key
The SEC already has AI on its radar. Other regulators will follow. As the SEC cases show this puts companies providing or using AI at risk regardless of whether there is any substance to the allegations. The fact that investigations are already underway, when so much about AI is still unclear and there is not even a standardised understanding of the term, only gives an idea of what companies can expect in the future. Therefore, companies should prepare accordingly if they are developing, integrating, or selling AI.
AI washing investigations are here to stay. They can threaten business models, trade secrets and reputations – and therefore endanger innovation. Companies need to prepare from the outset when describing their own capabilities and business activities related to AI. Preparation should not be limited to general measures to deal with an impending investigation. The central component of the preparation must be AI-specific: Companies should be able to document and provide evidence at short notice of what their AI application is capable of, when, and on what basis. This will enable them to respond promptly and rebut any allegations made. Companies should therefore prepare for possible investigations by being able to demonstrate that the application in question lives up to its claimed capabilities.
