A video claiming to show a humanoid robot attacking a woman in China has been circulating on social media, garnering thousands of worried comments. We’ve seen many posts speculating about the robot – wondering if it is gaining sentience and turning against humans – or dangerously malfunctioning due to its AI programming. Other users are labelling the robot “the Terminator” or suggesting that China is creating an “army of bots” for use in warfare.
In the video, a large crowd is standing behind a barrier watching several humanoid robots. One robot standing near the barrier appears to lunge forward into the crowd before being dragged back by security guards. While some posts claim that a woman was injured or physically “attacked”, the low quality of the clip makes it difficult see exactly what took place.
As the video is circulating without much further information or context, we decided to look deeper in order to fully understand the exact circumstances surrounding the incident.
The robot in question is a “humanoid agent AI avatar” made by the Chinese company, Unitree Robotics. A look at their website seems to confirm this, with the exact same robot (dressed in a traditional Chinese vest) on the home page.
The robot appears to be a “H1” model of their humanoid robots that was recently debuted at a popular annual live performance broadcast, Spring Festival Gala, on 28 January during Chinese New Year celebrations. However, the claim clip does not appear to be from the Spring Festival Gala as the setting is outdoors (rather than in an indoor broadcast venue). And, our initial reverse image searches of the video did not yield any hints as to its original source.
Instead, we used Chinese language search terms such as Unitree’s Chinese name, “宇树人形”, and the Chinese word for robot (“机器人”) to comb through image results on Google.
This led us to videos of the same robot in an outdoor setting – during a festival held in Taishan on 15 February. The security guards in both the claim video and new images from the Taishan Lantern Festival are dressed in the same uniform.
A more specific on Chinese social media platform Douyin finally led us to the original source of the claim video which also features different angles, much better image quality, and more written context about the situation.
Firstly, when examining the original sources, it does appear that the robot’s actions have sparked mixed discussion amongst some Chinese social media users – wondering if the robot is overexcited, clumsy or outright aggressive towards humans. There also does not appear to be any clarification or explanation from official sources (such as Unitree or festival organisers). However, the clearer videos also show that no actual physical contact was made and other reposts of the video have framed it as a comedic accident.
Some captions and comments also posit an explanation for the robot’s sudden lunge forward – that it simply tripped over the barriers that jut outward from the bottom. As other videos show, the bottom half of the barriers have a shelf that (unseen by the camera angle) could have hit the robot in the knees and caused it to tip forward.
We therefore give the claim that a humanoid robot “headbutted” or attacked a woman a rating of unproven based on the information available. While the video could be interpreted as an aggressive attempt at “headbutting,” there are other plausible explanations (for instance, an accidental fall) that have been left out from how the claim posts are being framed on English-language social media. Further, the low quality of the video in the claim posts also muddies the waters and causes uncertainty as to what actually happened.
The use of blurry clips and specific framing in order to evoke certain emotions or reactions is a tool is readily and frequently used by bad actors or content farmers to spread mis/disinformation. While the claim video might be cause for concern – especially given the global landscape of the international AI race – it is also important to base our reactions on the full picture.
More often than not, media such as videos or images are used to make claims and spread narratives online. Recognising how one single piece of media can be presented in multiple different angles is key to approaching all content with a critical eye – whether it is in the context of local events (such as the upcoming elections) or international crises (such as conflict in the Middle East)
And, while labelling a claim “unproven” can feel unsatisfactory, it is also an important reminder amidst a digital environment full of split ideologies and black/white binaries that is it okay (and even necessary) to not have a definite conclusion based on limited information.
