Securities Regulation Daily Wrap Up, RISK MANAGEMENT—Secretariat releases white paper on AI-focused securities class actions, (Feb 27, 2026)

By Rebecca E. Hoffman, J.D.

At this stage, with cases still ongoing, the paper noted, the data is not sufficient to calculate plaintiffs’ “win rate.”

Advisory firm Secretariat released a white paper Feb. 25 that examines AI-related securities class actions, which have been on the rise over the past five years, describing typical issues, median settlement totals, and preliminary outcomes.

“The past few years have seen a marked increase in securities class action litigation tied to artificial intelligence,” the report began, noting that numbers of such cases have risen from 3 in 2020 to 16 in 2025.

Themes. The authors observe that 65 percent of the 51 actions since 2020 include allegations of “AI washing,” where a company overstates or misrepresents its AI capabilities. “This pattern of exaggerating AI’s role—or failing to admit human involvement behind the scenes—is a core allegation in the majority of AI-related class actions,” they said.

Companies allegedly failing to tell shareholders about risks or limitations of AI is another emerging theme in these cases. “The SEC staff has warned companies not to suggest their AI tech is more ‘autonomous, scalable or mature’ than it really is,” the paper said, “and to disclose material AI risks such as data quality, bias, or cybersecurity vulnerabilities,” or risk a fraud action.

Less common, but occurring in 12 percent of cases, is the theme of algorithm failures which caused losses or setbacks that were not disclosed to shareholders in a timely manner. For example, in one case a credit underwriting model powered by AI was miscalculating the price of loans and required humans to jump in. “In these cases, … the complaint alleges that the company’s core technology underperformed or malfunctioned, causing operational damage … that was not accurately communicated to shareholders,” the authors explained.

The paper also observed that this trend has affected both companies with AI products and services and those with a more tangential involvement, such as companies making hardware or other infrastructure for AI purposes. It noted as well that the Second and Ninth Circuits host a majority of these cases, not unlike securities litigation in general.

Preliminary outcomes. Few of these cases have concluded at this point such that there are not many outcomes to analyze, the authors cautioned. But “[c]ontrary to speculation that courts might treat AI-related misstatements as a novel risk and perhaps be more lenient to investors at the pleading stage, AI cases have so far proven no more survivor-friendly than typical securities suits,” they said. The report observed that the median settlement amount in this category is $4 million, as compared to $13 million for securities class actions in general, but so few have settled at this point that it would not make sense to take a broad view of this discrepancy.

The authors conclude that although the allegations prevalent in AI-related actions mirror traditional cases, the AI context presents “subject matter that can be as complex as it is alluring.” Furthermore, if regulators are not acting to combat AI-related fraud, the “enforcement gap” will lead to more of these lawsuits. It is possible we may see guidance or rulemaking, which could “fuel more class actions by establishing yardsticks for falsehoods,” or could lead to improvements in disclosure and fewer actions, the report said. “For now, prudent companies are advised to treat AI-related statements with the same rigor as any other material assertion,” it recommended, adding that “what might seem like harmless hype … can become Exhibit A.”

LitigationEnforcement: AINews CorporateGovernance FraudManipulation InvestorEducation RiskManagement