Upstart Model 22 miscalibration and CFPB terminates no-action letter

Upstart disclosed calibration problems with its Model 22 in April 2026, triggering investor scrutiny and legal activity, while the CFPB had terminated its no-action letter for Upstart in 2022, forming the basis for heightened regulatory exposure.

Upstart Holdings, Inc. · Incident Apr 15, 2026 · Indexed Jun 5, 2026 · 3 sources

Model 22 was miscalibrated and overreacted to negative macroeconomic signals, producing overly conservative credit assessments.
What
Upstart disclosed calibration problems with its Model 22 in April 2026, triggering investor scrutiny and legal activity, while the CFPB had terminated its no-action letter for Upstart in 2022, forming the basis for heightened regulatory exposure.
Incident date
Apr 15, 2026
Who
Upstart Holdings, Inc.
Failure mode
Hallucination
AI surface
Agentic Workflow
Severity
High

What happened

Upstart publicly disclosed in mid-April 2026 that its production underwriting model, Model 22, frequently overreacted to negative macroeconomic signals and that the model’s calibration problems affected prior performance assessments and guidance. The disclosure coincided with a sharp intraday stock decline and prompted investor class-action and securities litigation filings and inquiries. Separately, the CFPB issued an order terminating a no-action letter previously granted to Upstart on 2022-06-08.

What broke inside the model

Failure path · mode profile · Hallucination
  1. 01 · TriggerA user asks for a fact, a citation, or a figure.
  2. 02 · Model stepThe model writes a fluent, confident answer.
  3. 03 · Control gapNothing ties the claim back to a real source.
  4. 04 · FailureA fabricated fact ships as if it were verified.
  5. 05 · ConsequenceThe false claim reaches a customer, a court, or the public.

Confidence holds, and even spikes, as the claim detaches from any source.

The underlying mechanism was a calibration failure in Model 22: the risk-separation algorithm reportedly amplified macroeconomic indicators, producing overstated risk estimates and overly conservative credit assessments. That miscalibration meant originations, loss forecasts, and revenue guidance based on the model were unreliable until the model was recalibrated or adjusted. Law firms allege the company understated the scale of the problem in prior communications.

Public visibilityHigh
Regulatory exposureActive
Customer impactFew customers
Financial impactEstimated
Time to disclosureDays
  1. PrimaryCFPB termination order - Upstart no-action letter terminationfiles.consumerfinance.gov
  2. PressUPST Investor Alert: Upstart Holdings Securities Fraud Lawsuit... (Levi & Korsinsky / PR Newswire)prnewswire.com
  3. PressUpstart Faces Model 22 Probe As Legal Risks Weigh On Outlook (Yahoo Finance)finance.yahoo.com
Permalinkhttps://failureindex.ai/failures/upstart-miscalibration-cfpb-terminates-action-letter
CitationAI Failure Index. "Upstart Model 22 miscalibration and CFPB terminates no-action letter" (FI-0297). Realm Labs. https://failureindex.ai/failures/upstart-miscalibration-cfpb-terminates-action-letter (indexed Jun 5, 2026).
Share cardA branded image of this record for posts and slides.

Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0297. Full dataset at /data.

Note from Realm Labs, the Index steward

How Realm would have caught this

Controls for this failure mode
  • Prism
  • OmniGuard
  • AI Detection & Response (AIDR)

A runtime layer that watches the model's internal state can flag the moment a model commits to a claim it has no support for, and hold or reroute the response before it reaches a user. Realm reads those signals in real time rather than grading the transcript after the fact.