Has Upstart had a publicly reported AI incident?
Yes. As of Oct 21, 2026, the register holds 3 publicly reported AI incidents at Upstart since January 2022. The most recent is "Upstart says its AI underwriting model over-reacted to macro signals, cutting approvals; investors sue" (Nov 4, 2025), status: court filing, severity 3 of 5. Last swept Oct 21, 2026 (16 queries).
| Date | Institution | Incident | AI? | Role | Category | Severity | Status |
|---|---|---|---|---|---|---|---|
| Nov 4, 2025 | Upstart | Upstart says its AI underwriting model over-reacted to macro signals, cutting approvals; investors sue | AI (stated) | Institution's own AI | Credit decisioning | 3 | Court filing |
| Mar 27, 2024 | Upstart | Independent monitor finds approval disparities for Black applicants in Upstart's AI lending model | AI (stated) | Governance finding | Bias finding | 2 | Confirmed by the institution |
| Jul 26, 2022 | Upstart | Investors' class action says Upstart misled them about its AI underwriting model | AI (stated) | Governance finding | Litigation | 2 | Court filing |
Nov 4, 2025 · Institution's own AI · severity 3/5
Upstart says its AI underwriting model over-reacted to macro signals, cutting approvals; investors sue
Upstart said on its third-quarter 2025 earnings call on 4 November 2025 that its AI underwriting model had tightened credit in response to macroeconomic signals over the summer, reducing approvals; CEO Dave Girouard said the model was "overreacting", according to American Banker. The conversion rate fell from 23.9% in the second quarter to 20.6%, and Upstart lowered its full-year 2025 revenue-from-fees outlook to about $946 million from about $990 million. On 7 April 2026 an investor filed a securities class action in the Northern District of California alleging that Upstart and four executives misled investors about the accuracy of the model, which the complaint calls Model 22. The case is pending, and we found no public statement from Upstart about it.
Mar 27, 2024 · Governance finding · severity 2/5
Independent monitor finds approval disparities for Black applicants in Upstart's AI lending model
Relman Colfax, the civil rights law firm acting as independent fair-lending monitor of Upstart's AI lending model, published its fourth and final report on 27 March 2024. The report, announced jointly by Upstart, the NAACP Legal Defense Fund and the Student Borrower Protection Center, found no evidence that the model's variables acted as close proxies for race or other protected classes and no pricing disparities above its thresholds. It did find approval disparities for Black applicants. Upstart adopted nearly all of the monitor's recommendations but declined a key one, on how to judge a less discriminatory alternative model, saying it would compromise accuracy. The monitorship ended at an impasse. No regulator action followed.
Jul 26, 2022 · Governance finding · severity 2/5
Investors' class action says Upstart misled them about its AI underwriting model
In re Upstart Holdings Securities Litigation, filed on 26 July 2022 in the US District Court for the Southern District of Ohio, alleges that Upstart and its executives misled investors, including about its AI underwriting model. In September 2023 the court let claims proceed over statements about the advantages of the AI model over FICO-based models and its ability to "quickly adjust" to changing economic conditions, among others. It certified an investor class in March 2025, allowed an amended complaint in September 2025, and on 30 September 2026 kept most claims against hedge fund Third Point. Upstart says the remaining claims are without merit. No ruling on liability has been made.
Not incidents: precautions and statements with no reported harm.
Entries come from the register of AI incidents at banks: events since January 2022, each sourced to two independent publishers or one official or court document, with the institution's response recorded. A sweep searches news, regulator, court and security sources for the institution by name; its date, query count and result are logged on the register's method section.
When AI goes wrong at a bank, the next morning's brief says so.
when a bank, fintech or regulator moves on AI, it is in the next morning's brief · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning