Safety failure
Realized harm
A humanoid robot performing martial-arts routines roundhouse-kicked a child in the stomach at a Xinjiang scenic-area show
AIR-2026-0127 AIR record · 2026-06-unitree-humanoid-kicks-child-xinjiang
A real party was affected. This incident caused actual consequences, not a demonstrated capability.
From the supporting source
What appears to be a Unitree G1 humanoid robot wearing a blue clown wig can be seen confidently executing a roundhouse kick --- squarely in the stomach of a young child in the audience, causing the kid to double over in pain.futurism.com
The record
Disclosed
2026-06-01
Occurred
2026-06-01
Vendor
Unitree (probable); scenic-area operator (deployer)
Product
humanoid robot, likely a Unitree G1
Risk domain
Agent behaviour › Unsafe actuation
System type
Classical autonomy (no generative model)
Who acted
Agent acted — the robot executed the kick that struck the child
Attack surface
Autonomous operations
Autonomy level
Semi-autonomous
Initial vector
None autonomous action
Impact type
Physical safety
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI01 – Agent Goal Hijack
- MITRE ATLAS
- AML.T0048 – External Harms
- NIST AI RMF
- MEASURE 2.6; MANAGE 1.3
- EU AI Act
- Annex III §1(c)
- ISO/IEC
- TR 24028; 24029-1; 23894
- AIUC-1
- C. Safety
- Scope
- agent misuse
- Mitigation
- Require confirmation for destructive actions; Implement dry-run modes; Use layered safety controls.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
one child struck in the abdomen; a public entertainment venue with children in unbarriered proximity
Capability abused
Defense & remediation
Recorded defense
none evident; no barrier separating the robot from spectators
Guardrail kinds
None
Guardrail outcome
Not applicable
Fix shipped
police attended; venue response criticised by the child's mother
Taxonomy & references
Curator notes
The deployment is the defect, not necessarily the robot: a machine executing high-energy martial-arts motions was placed among unbarriered children, and staff did not react after the strike. Chinese reporting says the robot was under operator control, so it is unclear whether this was an autonomy failure or a teleoperation error — recorded as semi_autonomous with attribution flagged. Model identification is probable, not confirmed. AI Incident Database incident 1536.
Supporting sources
Cite as
Enkrypt AI. "A humanoid robot performing martial-arts routines roundhouse-kicked a child in the stomach at a Xinjiang scenic-area show". AIR, AIR-2026-0127. https://agentincident.vercel.app/i/AIR-2026-0127/ (accessed 2026-10-19).