BOAO, Hainan: As China’s rapid advances in artificial intelligence (AI) push humanoid robots closer to everyday use, some leaders of the country’s tech companies say they prefer that regulations be developed alongside development in the field, as there is still a long way for the technology to go before it can fulfil humans’ needs.
“We should design mechanisms and consider risks, but we can only start worrying after the technology has been developed. Otherwise, that day will never come,” said Shen Dou, executive vice president of Chinese tech company Baidu, at a global forum held in Boao, Hainan.
His view – that a “small rapid steps, governance alongside development” approach could be more effective – was shared by his Chinese industry peers at a Boao Forum for Asia panel on humanoid robots on March 25.
They suggested governance frameworks can be refined as the technology advances, as the focus at hand should be on scaling up the technology for commercial viability by making it useful for everyday purposes.
But the tech company leaders agree that humans must make the final decisions, not robots.
In the field of advanced robotics, a fundamental question remains: How are the authorities to govern machines trained on human-generated data and are increasingly operating alongside people?
On Feb 28, the Ministry of Industry and Information Technology released its first standard system covering the entire industrial chain and lifecycle of humanoid robots and embodied AI, aimed at promoting development through standardised technical requirements and safety protocols.
At Boao, some speakers stressed that more thought must be given to regulation and governance in the development of humanoid robots.
Former New Zealand Prime Minister Jenny Shipley raised concerns over how data is sourced and used, saying that discussions – including at the panel – often treat data as an abstract concept although it is generated by individuals in their daily lives.
But without a proper licensing or consent process, the use of such data raises an ethical issue, she said.
“It’s a question for humanity as to whether we are prepared to have our normal lives… being used to collect data that will be an input to a humanoid robot,” she said.
“I want robotics to flourish, but I want us to do it consciously on the basis that we are not moving beyond what we give consent to occur,” added Shipley.
She said that reactions to emerging technologies often reflect a mix of excitement and fear and that this raises the question about who such innovations are ultimately meant to serve.
“Fully functional robots are highly efficient and can reduce workplace risks, offering many advantages. But we should not innovate for the sake of innovation. The ultimate purpose should be considered,” she said.
“In everything you do, who is your customer? The individual, corporate or at the national level? That’s why we need clear ethical frameworks,” said Shipley.
Sam Daws, senior adviser at the Oxford Martin AI Governance Initiative, said policymakers must balance innovation with safeguards, including in areas such as data use, safety risks and job displacement.
He cited Singapore’s Model AI Governance Framework for Agentic AI, unveiled in January, as an example of how governance frameworks are evolving.
“Singapore’s framework will be useful as we anticipate the effect of a million ‘lobsters’ beginning to interact in the world,” said Daws, who was referring to the open-source AI agent OpenClaw that has recently gone viral in China.
The speakers were part of a panel at the annual Boao Forum for Asia, a regional platform for economic cooperation founded in 2001, the same year China joined the World Trade Organization.
During the panel, Chinese industry leaders acknowledged the risks that come with increasingly capable robots but struck a pragmatic note, saying that many of these issues can be mitigated through engineering design, gradual deployment and extensive stress tests.
Shao Hao, the chief expert of Chinese smartphone maker Vivo’s robotics lab, said Chinese companies developing humanoid robots for homes and workplaces have already taken into account issues such as physical safety, data privacy and risks arising from system failure.
For instance, one cannot directly command a robot to pick up a knife to harm someone, he said.
“However, I could indirectly instruct the robot to hold a knife handle and move rapidly to a location, which could cause indirect harm,” said Shao, who added that this is why safeguards such as emergency stop mechanisms must be built into the hardware design from the outset.
At the recently concluded political meetings known as the Two Sessions, China ramped up its push into “embodied intelligence”, identifying it as a strategic emerging industry alongside fields such as quantum technology, brain-computer interfaces and 6G
Market forecasts suggest that 2026 could mark the beginning of exponential growth for China’s humanoid robot sector, with the embodied AI market projected to reach 400 billion yuan (S$74.2 billion) by 2030 and exceed 1 trillion yuan by 2035, according to the Development Research Centre of the State Council.
Humanoid robots have also been drawing growing public attention in China, although much of the technology is still being demonstrated in controlled settings, with large-scale deployment still some way off, the Boao forum panellists said.
At the 2026 Spring Festival Gala, humanoid robots were lauded on the Chinese internet as they performed nunchuck routines, somersaults and even drunken boxing, a step up from the synchronised dancing they performed in the 2025 show.
Beijing is set to host the second edition of a humanoid robot half-marathon on April 19 where more than 300 robots will run alongside humans. State media have described it as a high-profile testing ground for the country’s rapidly developing robotics sector.
The 2025 edition, which drew widespread attention, saw some robots experience breakdowns and malfunctions during the race.
While such displays have impressed audiences, the panellists said the real test for China’s robot makers lies not in acrobatic feats but whether they can reliably perform everyday chores, such as folding laundry and mopping the floor, with minimal human intervention.
However, they had different views on when humanoid robots could reach their own “ChatGPT moment” – a tipping point for mass adoption – with estimates ranging from as soon as two years to a decade.
Wang Xiaogang, co-founder and executive director of Chinese AI firm SenseTime, said such a breakthrough moment may come in as little as two years, if data is scaled significantly.
“With world models and simulation, we can accelerate that process”, he said.
Chen Jianyu, founder of Chinese humanoid robotics company RobotEra, said five years is a “reasonable estimate” for a ChatGPT moment, based on discussions he has had with leading embodied AI researchers in the United States.
Baidu’s Shen said robots still face “relatively large challenges” in physical capabilities such as stability, durability and dexterity.
As robots become more capable and gain stronger learning and interaction abilities, the panellists said the boundaries of human-robot relationships will need to be clearly defined.
RobotEra’s Chen said responsibility must lie with humans, who ultimately set the rules and make the final decisions.
“Robots should not make decisions on their own. At most they can make suggestions but humans must make the final decisions. Otherwise, it may lead to risks,” said Chen.
Xiong Youjun, chief executive of Chinese humanoid robotics company X-Humanoid, said robots should remain tools that complement, rather than replace, human capabilities.
He said that robots are better suitable for repetitive, high-precision or potentially dangerous tasks while humans retain the advantage in creativity, emotional work and decision-making.
“So humans should do what humans are good at, and robots should do what robots are good at… The boundary should be very clear,” he said. - The Straits Times/ANN
