SoftBank Robotics' Pepper robots reportedly suffered repeated mechanical and software failures

Media reports from mid-2021 alleged that SoftBank Robotics' Pepper humanoid robots experienced frequent mechanical errors, unplanned stops, failures to recognize people, and breakdowns while deployed in customer settings. The incidents were reported by multiple outlets and collected in the AI Incident Database, and customers were said to have declined renewals or resold units.

SoftBank Robotics (SoftBank Robotics Group Corp.) · Incident Jul 13, 2021 · Indexed Jun 10, 2026 · 4 sources

Reports allege perception and motor-control failures caused Pepper to stop, misidentify people, and break down during tasks.
What
Media reports from mid-2021 alleged that SoftBank Robotics' Pepper humanoid robots experienced frequent mechanical errors, unplanned stops, failures to recognize people, and breakdowns while deployed in customer settings.
Incident date
Jul 13, 2021
Who
SoftBank Robotics (SoftBank Robotics Group Corp.)
Failure mode
Agentic Action Error
AI surface
Autonomous System
Severity
Medium

What happened

Media accounts in mid-2021 reported that deployed Pepper robots frequently experienced mechanical errors, unplanned stoppages, failures to recognize previously-met people, and breakdowns during practice or live use in settings such as nursing homes, bank lobbies, and a funeral service. The reporting said some customers declined to renew contracts and listed used Pepper units for sale online. These outcomes were reported by independent news outlets and recorded in the AI Incident Database; the allegations have not, in the sources cited, been presented as court-adjudicated findings.

What broke inside the model

Failure path · mode profile · Agentic Action Error
  1. 01 · TriggerAn agent plans a multi-step task.
  2. 02 · Model stepIt chooses a wrong or destructive action.
  3. 03 · Control gapNo confirmation gate guards the write.
  4. 04 · FailureThe action commits to a system of record.
  5. 05 · ConsequenceData is changed or destroyed irreversibly.

A wrong action commits, and the step is written before anything can stop it.

Reporting attributed the problems to a lack of robustness in perception and motor-control capabilities and to hardware reliability issues that produced repeated physical errors and breakdowns. Observers quoted in the press suggested the robot’s locomotion, manipulation, and recognition systems did not reliably meet customer expectations in real-world deployments.

Public visibilityHigh
Regulatory exposurePossible
Customer impactMany customers
Financial impactEstimated
Time to disclosureMonths
  1. PrimaryIncident 152: SoftBank's Humanoid Robot, Pepper, Reportedly Frequently Made Errors, Prompting Dismissalincidentdatabase.ai
  2. PressSoftbank’s Hyped Robot Keeps Failing at Its Jobsinc.com
  3. PressHumanoid Robot Keeps Getting Fired From His Jobs (Wall Street Journal)wsj.com
  4. PressEXCLUSIVE SoftBank shrinks robotics business, stops Pepper production, sources sayreuters.com
Permalinkhttps://failureindex.ai/failures/softbank-robotics-pepper-robots-reportedly-suffered
CitationAI Failure Index. "SoftBank Robotics' Pepper robots reportedly suffered repeated mechanical and software failures" (FI-0397). Realm Labs. https://failureindex.ai/failures/softbank-robotics-pepper-robots-reportedly-suffered (indexed Jun 10, 2026).
Share cardA branded image of this record for posts and slides.

Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0397. Full dataset at /data.

Note from Realm Labs, the Index steward

How Realm fits

Controls for this failure mode
  • Prism
  • OmniGuard
  • AgentRealm

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.