One of the interesting features of the rise of AI has been the advent of “AI-and” businesses – that is, businesses whose strategy is to apply AI tools to traditional business models. When “AI-and” business results fall short, securities litigation has sometimes followed. In the latest example of this kind of litigation, earlier this week a plaintiff shareholder filed a securities suit against Upstart Holdings, a company whose business model involves applying AI tools to traditional credit rating and lending services, after the results from the company’s AI-updated credit rating tool disappointed investors. A copy of the new Upstart Holdings complaint can be found here.

Background

Upstart is a cloud-based artificial intelligence (AI) lending platform. The company uses its own AI models to quantify loan risk. The company claims that its credit rating approach generally leads to higher loan and credit approvals and lower interest rates relative to traditional lending practices, providing more predictable returns to its capital partners (banks, credit unions, and institutional investors). The company periodically updates its AI models to try to improve its approval process.

In May 2025, the company launched the latest iteration of its AI-based approval model. The updated model was called “Model 22.” The company touted the accuracy of Model 22, claiming that it would increase loan approval rates and, accordingly, the company’s revenue. Indeed, in both May 2025 and August 2025, the company raised its annual revenue guidance.

On November 4, 2025, the company released its financial results for 3Q25. The company’s reported quarterly revenues missed its previously issued guidance as well as consensus estimates, and the company also lowered its guidance for 4Q25 and full year 2025.

In an earnings call that same day, the company blamed its disappointing 3Q results on Model 22, which the company said had “overreacted” to macroeconomic signals during the quarter, lowering approvals and conversion rates. The company also acknowledged that it had “knowingly” calibrated the AI model to be “more conservative on the credit side,” and that negative impacts of Model 22’s “over-responsiveness” to macroeconomic signals would continue to negatively impact the company’s 4Q and full year revenues. The company’s share price fell approximately 10% on this news.

The Lawsuit

On April 7, 2026, a plaintiff shareholder filed a securities class action lawsuit in the Northern District of California against Upstart and certain of its executives. The complaint purports to be filed on behalf of investors who purchased the company’s securities between May 14, 2025, and November 4, 2025.

The complaint alleges that during the class period the defendants failed to disclose that: “(i) Model 22 frequently overreacted to negative macroeconomic signals in performing its risk-separation processes; (ii) accordingly, Model 22’s overall accuracy and propensity to increase loan approval rates was overstated; (iii) Model 22’s overly conservative assessment of credit and macroeconomic conditions was having a significantly negative impact on Upstart’s revenue results, rendering the Company’s previously issue FY 2025 revenue guidance unreliable and/or unrealistic; and (iv) as a result, Defendants’ public statements were materially false and  misleading at all relevant times.”

The complaint alleges that the defendants violated Section 10(b) and 20(a) of the Securities Exchange Act of 1934 and Rule 10b-5 thereunder. The complaint seeks to recover damages on behalf of the class.

Discussion

As I noted at the outset, this company represents an example of the new AI adaptive approach to a traditional business model, in which the company pursues an “AI-and” strategy. Here, the company basically provide a credit rating and loan approval service. The difference is that the company claimed to provide superior services through its application of AI tools. It sought to differentiate itself by portraying its approach as an “AI-and-credit rating” model.

As more and more companies seek to differentiate themselves by trying to characterize their strategy as an “AI-and” business model, problems can arise when the supposed advantages of the  AI-enhanced business model falls short of expectations, as happened with this company. Some readers may accurately perceive the allegations that can follow in the wake of the shortfall as “AI washing”-type allegations. The important point here is that “AI-and” businesses can be particularly susceptible to these kinds of allegations, as their AI-differentiated business approach is built on the premise that the AI-based approach will produce superior results – a premise that potentially sets the company up for a fall (and possibly for securities litigation) if and when the AI-based approach falls short of expectations.

I have to say that after a lifetime of watching how various circumstances can lead to securities litigation, this case surprised me a little bit. My decades of experience caused me to expect that this company’s business model might lead to problems if its AI-based model proved to be too liberal in granting credit approval to marginal credit risks. Ironically (at least in my view) this company ran into problems not because its model was too liberal, but rather when the company’s revised AI model proved to be too conservative. Honestly, it is not usually business models that prove to be too conservative that cause companies problems.

In any event, this case joins the growing list of AI-related securities lawsuits that have been filed so far this year. Based on my count, this case is the eighth AI-related securities suit to be filed so far in 2026. It seems likely that when we reach year-end and we tally up all of the securities suits filed during the year, the AI-related securities suits will represent a significant part of the year’s total securities lawsuit filings.