China has blocked Meta’s acquisition of AI startup Manus, the top economic planning body said Monday, after a regulatory review that reportedly also saw Beijing restrict two co-founders from leaving the country.

Facebook owner Meta had agreed to acquire Manus, an artificial intelligence agent created by a company founded in China but now based in Singapore, the two firms said in December.

Analysts however had warned then the deal could fall foul of regulators at a time of fierce technological rivalry between Washington and Beijing.

The Financial Times reported last month that China had restricted two Manus co-founders from leaving the country, citing three people with knowledge of the matter.

Chief executive Xiao Hong and chief scientist Ji Yichao, who are usually based in Singapore, were reportedly summoned to a meeting in Beijing in March and told they were not allowed to leave China because of a regulatory review of the Meta acquisition.

Beijing’s National Development and Reform Commission said in a statement on Monday that it will “prohibit the foreign investment in the acquisition of the Manus project” and “requires the parties involved to withdraw the acquisition transaction”, without naming Meta.

It added that this was done “in accordance with laws and regulations”.

AFP has contacted Manus and Meta for comment.

Meta said in December that the deal — the financial details of which were not disclosed — would “bring a leading agent to billions of people and unlock opportunities for businesses across our products”.

Bloomberg Intelligence analysts said the purchase was likely aimed at expanding Meta’s AI agent task capabilities, and that it could be worth more than US$2 billion.

Manus, created by startup Butterfly Effect, can sift through and summarise resumes or create a stock analysis website, according to its website.

The White House on Thursday accused Chinese entities of a massive effort to steal US artificial intelligence technology and vowed to take action to prevent the alleged theft.

“The US has evidence that foreign entities, primarily in China, are running industrial-scale distillation campaigns to steal American AI,” White House science and technology chief Michael Kratsios said in a post on X.

“We will be taking action to protect American innovation.”

Distillation is a common practice within AI development, often used by companies to create cheaper, smaller versions of their own models.

In February, US AI developer Anthropic accused three Chinese firms, DeepSeek, Moonshot AI and MiniMax, of running campaigns to illicitly extract capabilities from its Claude chatbot, describing it as intellectual property theft.

That same month, ChatGPT creator OpenAI sent a letter to US legislators accusing DeepSeek of using distillation techniques amid “ongoing efforts to free-ride on the capabilities developed by OpenAI and other US frontier labs.”

Kratsios did not name any specific foreign entities in his post but said they “are using tens of thousands of proxies and jailbreaking techniques in coordinated campaigns to systematically extract American breakthroughs.”

The accusations come ahead of a planned May 14 summit in Beijing between US President Donald Trump and his Chinese counterpart Xi Jinping.

A humanoid robot competing against flesh-and-blood runners broke the world record at a Beijing half marathon on Sunday, showcasing the rapid technological advancement achieved by Chinese makers.

Spectators lined the roads in Yizhuang in the capital’s south to watch the machines and their human rivals race, each group in a separate lane to avoid accidents or collisions.

Some of the robots were highly agile, moving like famous runners such as Usain Bolt, while others had more basic capabilities.

The winning humanoid, equipped with an autonomous navigation system and running for Chinese smartphone maker Honor, completed the roughly 21-kilometre (13-mile) course in 50 minutes and 26 seconds, at an average speed of about 25 kilometres per hour, according to state broadcaster CCTV.

That was far faster than the top human in Sunday’s race, while also surpassing the current men’s world record of 57:20 held by Ugandan runner Jacob Kiplimo.

The result represented spectacular progress from last year, when robot-runners fell repeatedly and the best took more than two hours and 40 minutes to finish.

The number of humanoid entries jumped from around 20 last year to more than 100, according to organisers, a sign of the sector’s growing popularity.

‘Pretty cool’

Han Chenyu, a 25-year-old student who watched the race from behind a safety barrier, barely had time to take out her phone and snap a picture of the leading robot as it whizzed past.

She told AFP she was enthusiastic about such leaps in technology and thought the event was “pretty cool”.

But, she added, “as someone who works for a living, I’m a little worried about it sometimes. I feel like technology is advancing so fast that it might start affecting people’s jobs”, particularly with artificial intelligence growing increasingly sophisticated.

Humanoid robots have become a common sight in China in recent years, in the media as well as in public spaces.

Xie Lei, 41, who watched Sunday’s race with his family, said robots could “become part of our daily lives” within several years, potentially used for “things like housework, elderly companionship or basic caregiving” or “dangerous jobs, even firefighting”.

The humanoid half marathon aims to encourage innovation and popularise the technologies used in creating and operating such machines.

In a sign of the industry’s strength, investment in robotics and so-called embodied AI amounted to 73.5 billion yuan ($10.8 billion) in China in 2025, according to a study by a government agency.

“For thousands of years, humans have been at the top on planet Earth. But now, look at robots. Just in terms of autonomous navigation, at least in this specific sport event, they’re already starting to surpass us,” Xie said.

“On one hand, it does make you feel a little bit sad for humanity. But at the same time, technology, especially in recent years, has given us so much imagination.”

Erotic chatbots, video-synced and voice-activated devices mesmerised visitors at a sex toy expo in Shanghai this weekend, as China’s adult product firms join the global AI craze.

China is the world’s largest producer of sex toys, and the country’s entrepreneurs have fully embraced AI tools in other sectors.

Though some businesses at the trade show said they were wary of legal risks posed by machine-generated sexual content, others were keen to advertise their enhanced wares when AFP visited on Friday.

Banners scattered across the large exhibition hall advertised a Guangzhou-based company’s “AI character dating” Luvmazer app, which turns conversations with virtual partners into vibrator pulses.

“One sentence can make you shiver,” the banners promised.

A life-size, cyberpunk-inspired silicon doll with metal joints lay in a large vitrine at the Cydoll booth, which factory manager Zhou Yuanqing said was a prototype designed to display “natural” emotions and speech.

“People nowadays don’t go out to drink or meet their friends, and they might prefer to play games on their phones or computers on their own… but they still need companionship,” Zhou told AFP.

Multiple sex toy companies showed off apps that use machine learning to interpret adult videos and translate them into pressure, speed and pulse pattern changes — features that were rare novelties just a few years ago.

“Everyone has the video syncing feature now,” an employee at teledildonics firm Amorlink told AFP at the company’s stand, which showcased vacuum cups equipped with powerful computing chips.

A two-in-one suction vibrator exhibited by domestic condom giant Jissbon meanwhile featured long-distance remote control capabilities, a roster of virtual “boyfriend” personas and the ability to match frequency and intensity to environmental noise levels.

Regulatory risks

Multiple businesses offered AI agents for marketing and operating offline adult stores, while others touted tech solutions for brands trying to make their own “smart” toys.

A poster for Hong Kong-based metaXsire, which did not have a booth at the expo, boasted an “adult image and video generator” and dirty talk in more than 80 languages.

The company’s website says its app can swap the faces of celebrities or personal acquaintances onto pornographic videos which are then synced to its toys.

The terms and conditions of the app forbid customers to use it to shame or harass others, but did not detail how the company would ensure consent for the faceswap feature.

But multiple exhibitors present on Friday said they were cautious about mixing AI and adult video content, due to legal and privacy concerns.

Pornography is technically illegal in mainland China, and most adult video sites are blocked by the country’s Great Firewall and cannot be accessed without specialised VPN software.

Sam Xie, the founder of Shanghai-based adult toy maker Magic Motion, said his products were compatible with AI agents but he had to be selective when choosing software developers to partner with.

“We have to be very careful, or there could be all sorts of problems, and we could get reported by consumers,” Xie told AFP.

By Katie Forster, with Luna Lin in Beijing

For weeks now, the global tech industry has been waiting for a major artificial intelligence launch from DeepSeek, seen as a benchmark for China’s progress in the fast-moving field.

More than a year has passed since the startup put Chinese AI on the map in early 2025 with a low-cost chatbot that performed at a similar level to US rivals.

But despite reports and rumours about its imminent release, DeepSeek’s next-generation “V4” model is nowhere in sight.

Speculation is also swirling over the geopolitical implications of which computer chips were chosen to train and power the new system: world-leading US designs or made-in-China alternatives that the country is racing to develop.

“It’s important to know because at one level, it is a signal of China’s AI self-sufficiency trajectory,” Wei Sun, principal AI analyst at Counterpoint Research, told AFP.

Tech news outlet The Information reported last week that V4 can be run on the latest chips made by China’s Huawei.

Such a shift would mark a milestone for China in its bid to beat US restrictions on the export of top-of-the-range AI chips from Californian titan Nvidia to the country.

The report cited five people with direct knowledge of large orders for Huawei chips, made in preparation for the DeepSeek launch by tech giants including Alibaba, ByteDance and Tencent.

AFP contacted DeepSeek, Huawei, Alibaba, ByteDance and Tencent but none were able to comment.

‘Wake-up call’

DeepSeek started life in 2023 as a side project of a hedge fund that had access to a cache of powerful Nvidia processors.

It shot to attention in January 2025 with its R1 deep-reasoning chatbot, which sent US tech shares tumbling with President Donald Trump calling it a “wake-up call” for American firms.

R1 was based on DeepSeek’s last major AI model, V3, which was released in December 2024.

The company’s affordable, customisable AI tools have been widely adopted in China, and are also popular in emerging markets such as Southeast Asia and the Middle East.

Stephen Wu, founder of the Carthage Capital fund, told AFP that V4 — said to be multimodal, meaning it can generate text, pictures and video — could again shock US tech valuations.

“I expect the upcoming DeepSeek V4 release will not just be a software update; it will be a highly capable, open-source model that handles massive context windows at a fraction of the cost,” he predicted.

But DeepSeek’s reputation as a company at the frontier of AI technology is also at stake.

Its models previously relied on Nvidia chips, so a move to collaborate with domestic chipmakers would require “substantial re-engineering”, Wei said.

“That transition can slow development cycles and introduce performance trade-offs, especially for V4, a model expected to be state-of-the-art.”

Training vs inference

The US cites national security concerns as the reason for its export ban on Nvidia’s most powerful AI processors to China.

“The ongoing wait for DeepSeek V4 points to friction in scaling advanced models without unrestricted access to top-tier Nvidia hardware,” Wu said.

But some reports allege that DeepSeek skirted the ban to train V4 using thousands of Nvidia’s top-end Blackwell chips, dismantled in third countries and smuggled to China.

Training AI models requires huge amounts of computing power — much more than for processing generative AI queries, which is known as inference.

AFP has contacted DeepSeek for comment. Nvidia did not respond to a comment request but told The Information it had not seen evidence of this and “such smuggling seems farfetched”.

Another Chinese AI startup, Zhipu, in January unveiled an image generator that it said had been entirely trained on Huawei chips.

And Alibaba said this week it would open a new data centre for AI training and inference in southern China, powered by 10,000 of its own chips and operated by China Telecom.

As for DeepSeek, “if they have successfully trained V4 entirely on Huawei silicon, it signals a material shift in the geopolitical tech landscape”, Wu said.