Sainsbury's customer wrongly ejected after facial recognition error
A customer at a Sainsbury's store in Elephant and Castle was misidentified as a known offender by the Facewatch facial recognition system. Although the system issued an alert, the incident was categorized as a human error where staff approached the wrong individual. Sainsbury's apologized and provided a voucher to the affected customer.
The error was made at the second stage of human verification.
Key facts
- What
- A customer at a Sainsbury's store in Elephant and Castle was misidentified as a known offender by the Facewatch facial recognition system.
- Incident date
- Jan 27, 2026
- Who
- Sainsbury's
- Failure mode
- Hallucination
- AI surface
- Computer Vision
- Severity
- Medium
What happened
Warren Rajah was shopping at a south London Sainsbury's on January 27, 2026, when staff approached him and asked him to leave the store without explanation. He was later informed that facial recognition technology had identified him as an offender. Sainsbury's apologized for the mistake and provided Mr Rajah with a £75 shopping voucher.
What broke inside the model
- 01 · TriggerA user asks for a fact, a citation, or a figure.
- 02 · Model stepThe model writes a fluent, confident answer.
- 03 · Control gapNothing ties the claim back to a real source.
- 04 · FailureA fabricated fact ships as if it were verified.
- 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 failure occurred during the human-in-the-loop verification stage of the process. While the Facewatch system generated an alert, store staff incorrectly identified the customer as the person flagged in the system. The failure was a breakdown in manual verification rather than a software-only error.
What it cost
Sources
- PressFacial recognition error: Customer misidentified by Sainsbury'sbbc.co.uk
- PressSainsbury’s shopper felt like a ‘criminal’ after facial recognition errorthe-independent.com
Cite this entry
https://failureindex.ai/failures/sainsbury-customer-wrongly-ejected-facial-recognitionAI Failure Index. "Sainsbury's customer wrongly ejected after facial recognition error" (FI-0323). Realm Labs. https://failureindex.ai/failures/sainsbury-customer-wrongly-ejected-facial-recognition (indexed Jun 8, 2026).Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0323. Full dataset at /data.
Note from Realm Labs, the Index steward
How Realm fits
- Prism
- OmniGuard
- AI Detection & Response (AIDR)
This entry sits in the index's predictive wing: a system that scores, ranks, perceives, or steers rather than generates. Realm's runtime layer is built for the generative and agentic systems now moving into these same decision seats, where it watches a model's internal state and holds an unsupported claim or an unchecked action before it commits. The control gap on this record, an automated decision that reached people with no runtime check in front of it, is the same gap. The index keeps predictive failures on the record because the pattern carries straight into the systems shipping today.