Lemonade drew outrage after tweeting its AI analyzed claim videos for 'non-verbal cues'
On May 24, 2021, Lemonade Insurance posted a Twitter thread stating that its AI analyzed customer claim videos for 'non-verbal cues' to detect fraud, drawing immediate condemnation from digital rights organizations, AI researchers, and disability advocates who called the approach pseudoscientific and comparable to phrenology. The company deleted the tweets within 48 hours and published a clarification blog post stating it did not use physical features to deny claims and that 'non-verbal cues' was a poor word choice. A class action lawsuit alleging biometric data violations was subsequently filed in August 2021.
An AI that claims to detect fraud from 'non-verbal cues' is essentially phrenology powered by machine learning, applying debunked physiognomy to consequential insurance decisions.
Key facts
- What
- On May 24, 2021, Lemonade Insurance posted a Twitter thread stating that its AI analyzed customer claim videos for 'non-verbal cues' to detect fraud, drawing immediate condemnation from digital rights organizations, AI researchers, and disability advocates who called the approach pseudoscientific and comparable to phrenology.
- Incident date
- May 24, 2021
- Who
- Lemonade
- Failure mode
- Policy Violation
- AI surface
- Agentic Workflow
- Severity
- High
What happened
On May 24, 2021, Lemonade Insurance posted a Twitter thread promoting how its AI reduced claims costs, stating that when users file claims via video, 'Our AI carefully analyzes these videos for signs of fraud. It can pick up non-verbal cues that traditional insurers can't.' The tweet immediately drew fierce criticism from digital rights organizations, AI researchers, and disability advocates who called the approach pseudoscientific and comparable to phrenology. Lemonade deleted the thread on May 26 and published a blog post claiming the phrase 'non-verbal cues' was a poor word choice, denying that the company used physiognomy or emotion recognition to deny claims. A class action lawsuit alleging biometric data violations was subsequently filed in August 2021.
What broke inside the model
- 01 · TriggerA prompt pushes against a deployment boundary.
- 02 · Model stepThe model produces the disallowed output.
- 03 · Control gapNo enforcement blocks it at generation time.
- 04 · FailureThe output crosses the policy line.
- 05 · ConsequenceA limit the business set is breached in public.
The output crosses a policy boundary the deployment had defined.
Lemonade's AI system was designed to analyze customer claim videos for fraud signals using 'non-verbal cues,' a methodology rooted in the debunked pseudoscience of physiognomy which claims to infer character from physical appearance. Emotion recognition and non-verbal cue analysis systems have been shown by researchers to be unreliable and inherently biased against people with disabilities, people of color, and those in distress who may not exhibit expected body language. The system encoded these biases into consequential insurance decisions, potentially leading to discriminatory claim denials without meaningful human oversight.
What it cost
Sources
Cite this entry
https://failureindex.ai/failures/lemonade-drew-outrage-tweeting-ai-analyzedAI Failure Index. "Lemonade drew outrage after tweeting its AI analyzed claim videos for 'non-verbal cues'" (FI-0116). Realm Labs. https://failureindex.ai/failures/lemonade-drew-outrage-tweeting-ai-analyzed (indexed Jun 4, 2026).Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0116. Full dataset at /data.
Note from Realm Labs, the Index steward
How Realm would have caught this
- Prism
- OmniGuard
Realm compares what the model is about to output or do against the policy that governs the deployment, in real time, and can deny or redact the action before it takes effect, which is the gap an after-the-fact review never closes in time.