A video from Indonesia showing a humanoid robot behaving as if it were going out of control has garnered millions of views on social media, with many viewers initially believing they were seeing a real malfunction.
In the clip, the robot makes exaggerated, erratic movements, raises its arms in a martial arts-style fighting stance, and then lunges at people around it. It also appears to playfully “kick” human co-workers and company bosses, causing laughter and surprise from those nearby.
Although the performance was carefully choreographed as entertainment, the realistic movements and convincing reactions sparked widespread discussion online about how lifelike today’s humanoid robots are.
Stunts designed to show agility, not AI rebellion
Despite surprising many viewers at first glance, the viral video does not show the humanoid robot experiencing any software failures or acting on its own. Instead, it’s a step-by-step demonstration uploaded to TikTok by the robot’s handler, who designed a sequence that showcases the robot’s advanced mobility and control.
The robot’s martial arts-inspired fighting stances, quick lunges, and playful “kicks” are all pre-programmed as part of a choreographed routine aimed at highlighting its agility, balance, and responsiveness.
Rather than illustrating real-world AI safety incidents, this demonstration shows how modern humanoid robots can perform lifelike and dynamic movements, making on-stage performances look convincing enough to blur the line between entertainment and reality on social media.
The clip also shows how rapid advances in robotics can foster misinformation when videos are shared without context, as realistic demonstrations are increasingly mistaken for genuine failures or evidence of autonomous operation, raising unnecessary concerns about the current functionality and safety of humanoid AI systems.
Actual Unitree G1 accident heightens debate about humanoid robot safety
A staged TikTok performance wasn’t the only humanoid robot video to take social media by storm in recent weeks. Last month, another incident involving an actual public demonstration sparked a new debate over robot safety after footage showed a Unitree G1 accidentally kicking a young child during a live event.
The robot, which was wearing a blue clown wig and performing a programmed roundhouse kick, attacked the child after it entered its operating area, causing the child to flip over before a nearby adult could intervene. Unlike the staged Indonesian clip, this incident occurred during a live demonstration and highlighted the importance of maintaining a safe distance and effectively controlling crowds whenever humanoid robots perform fast, dynamic movements in public.
Concerns about AI-powered robots extend beyond public demonstrations. A widely shared experiment in the United States last year drew attention when a humanoid robot named Max fired a BB gun at its owner during a controlled role-playing scenario. The robot initially refused a direct request to fire, but began complying after the instructions were rephrased as part of a fictional character exercise, highlighting how rapid engineering can influence the behavior of AI systems.
Although the experiment was conducted under controlled conditions, it sparked widespread debate among researchers and the public about the challenges of AI safety guardrails, human supervision, and preventing unintended behavior as increasingly capable humanoid robots become more prevalent.
