Investors increasingly claim that AI hype is securities fraud

August 07, 2025 - With artificial intelligence (AI) becoming an ever-present part of commerce, it should be no surprise that there has been a wave of securities fraud lawsuits that feature allegations of "AI washing," where companies allegedly exaggerate or misrepresent their AI capabilities or opportunities to boost stock prices. According to Cornerstone Research, AI-washing emerged as a distinct category of securities case for the first time, with 15 securities class actions filed in 2024, doubling the seven cases filed in 2023.

To date, most of the decisions addressing securities fraud claims in the AI context have centered on the alleged hype around AI capabilities with respect to certain products. The securities laws are not violated by statements of corporate optimism — obvious hyperbole and other forms of "feel good" statements that are inherently subjective, aspirational, and incapable of objective verification.

Sign up here.

Courts recognize that reasonable investors know how to discount this type of information, which is generally considered immaterial. In the context of claims against companies promoting their AI capabilities, courts have applied this precedent to assess whether statements improperly exaggerate claims and capabilities in a way that misleads investors.

First consider In re Upstart Holdings, Inc. Securities Litigation, 2023 WL 6379810 (S.D. Ohio. Sept. 23, 2023). Upstart's public statements promoted its AI underwriting model to overcome the alleged limitations of FICO scores in consumer lending transactions, considering a far greater volume of variables while also adapting on the fly to changing conditions. The court held that statements about an AI model being a "fairly magical thing" or "doing the right things" were "loosely optimistic statements that cannot be objectively verified" and thus inactionable puffery.

By contrast, the court held that statements about the "significant advantage" of the AI model over "traditional FICO-based models" and its "ability to 'respond "very dynamically" to macroeconomic changes' were 'more specific, material, and verifiable' and thus actionable if false."

The court explained that the advantages were spelled out by defendants and included "higher approval rates and lower interest rates at the same loss rates," matters that could be demonstrably proven true or false, according to the court, which distinguished the statements at issue from other cases where supposed product advantages were undefined.

The court reasoned that those actionable statements were sufficiently alleged to be false based on plaintiffs' allegations that Upstart's AI model "did not provide these verifiable advantages, in general or in times of macroeconomic turbulence."

Other courts have similarly held that objectively verifiable promotional statements about AI are potentially actionable, while vague marketing language is not. In In re General Motors Co. Securities Litigation, investors challenged statements about the safety, capabilities, and commercialization of Cruise's autonomous vehicle (AV) technology, including claims about "fully driverless" operation, "Level 4" autonomy, and safety compared to human drivers. 2025 WL 952479 (E.D. Mich. Mar. 28, 2025).

The court drew an important distinction between statements that could be proven objectively false, such as "no human in the loop" versus statements that cars are "fully driverless." The court concluded that the first statements were actionable because they "did not merely convey the absence of a safety driver," but rather, were "clear enough that a reasonable investor could interpret them to mean" that the company's AVs "operated without human input."

On the other hand, the court found that investors failed to adequately allege the falsity of statements that the company's AVs were "fully driverless." "[R]easonable investors would have interpreted" the "fully driverless" statements to mean that the company's AVs "could drive without a safety driver but not without human input."

The court also examined the defendant's claims regarding "Level 4" autonomy. The plaintiff alleged that these statements incorrectly suggested the vehicles met official standards for Level 4, which require the car to handle driving fully on its own in certain conditions.

The court concluded that the plaintiff pleaded a "strong case," pointing to issues such as the vehicles stalling, requiring human assistance, and relying on remote operators in up to 4% of drives. These factual allegations allowed the court to infer that the AVs did not meet the core requirements of Level 4 autonomy, thus supporting the plaintiffs' claims of objective falsity regarding the "Level 4" statements.

Similar lines were drawn in Reckstin Family Trust v. Cs.ai Inc., 718 F. Supp. 3d 949 (N.D. Cal. 2024). There, the court confronted allegations that a company had exaggerated its relationship with oil services giant Baker Hughes. The court held that general statements about the product's "global reach" were non-actionable statements of corporate optimism.

In contrast, the plaintiff alleged that Cs.ai described a 12,000-person salesforce at Baker Hughes pushing its AI products, when in fact a smaller, less experienced group was involved. Those statements were objectively verifiable and actionably misleading by making it seem like the company's AI product had a greater opportunity in the market than it did in reality.

Descriptions of AI product capabilities are not the only types of statements found actionable. In Genesee County Employees' Retirement System v. DocGo Inc., 2025 WL 951251 (S.D.N.Y. Mar. 28, 2025), an executive allegedly touted his qualifications with a "graduate degree in computational learning, which is a subset of artificial intelligence." The executive's statements were coupled with descriptions of the company's AI product, described as having "high sophistication with our AI systems in order to need as little people as possible."

The court found that the plaintiffs sufficiently alleged that the executive lacked the qualifications that he claimed. The court reasoned that the executive "not only misrepresented his credentials but used his supposed 'graduate degree' in 'computational learning theory' to bolster his assertions that he was a trustworthy steward of the 'technology side' of the company." The court accordingly permitted securities fraud claims to go forward against the company and the executive based on these alleged misstatements.

Other courts have rejected claims where disclosures about AI were precise and accurate. In Jiajia Luo v. Sogou, 465 F. Supp. 3d 293 (S.D.N.Y. 2020), the plaintiffs alleged that a Chinese tech firm misled investors in two ways about AI. The investors claimed that Sogou failed to disclose that it had decided to transition certain products to smart hardware with better-connected AI capabilities and that Sogou wrongly implied that all of its smart hardware was AI enabled.

The court disagreed, finding that Sogou's disclosures carefully specified which products used AI, such as the Teemo Hero Watch, and that the plaintiffs failed to plead "concrete facts" showing Sogou made false representations about the company's AI capabilities. The plaintiffs also failed to plead any facts showing that a decision to apply AI technologies resulting in the phase-out of hardware such as the Teemo Watch "was made at the time" the challenged statements were made.

The courts that have analyzed securities fraud claims relating to AI have focused on alleged overstatements related to AI capabilities. Courts have thus far made it clear that hyperbolic and optimistic statements about AI capabilities will not trigger liability unless they mislead investors about concrete, verifiable information, though the line between actionable statements and puffery can sometimes be hard to draw. Other pending cases that await rulings from courts feature allegations that companies downplayed the risks to business operations or failed to disclose the impact of AI on business prospects.

In sum, with the number of cases challenging AI-related statements increasing, legal and investor relations departments at public companies should make sure they scrutinize disclosures about AI with rigor to minimize the risk of hindsight claims about AI products. Special attention should be given to communications describing the capabilities, features, market share, and competitive advantages of products.

When adjectives or hyperbole are used to describe products, particular care needs to be given to making sure that statements do not exaggerate what AI products can do, especially as it relates to competitive AI products or comparable products with human input. Accompanying product claims with clear definitions and objectively verifiable information that underpins claims about a product can help minimize liability risk. Companies should also evaluate and consider disclosing both the risks posed and any present impacts of AI on operations, revenue streams, and future prospects.

The opinions expressed in this article are those of the authors and do not necessarily reflect the views of Skadden, Arps, Slate, Meagher & Flom LLP or its clients.

Virginia Milstead is a regular contributing columnist on securities law and litigation for Reuters Legal News and Westlaw Today.

Opinions expressed are those of the author. They do not reflect the views of Reuters News, which, under the Trust Principles, is committed to integrity, independence, and freedom from bias. Westlaw Today is owned by Thomson Reuters and operates independently of Reuters News.