Hong Kong scholar Joseph Li has had to rethink the design of his courses almost every semester over the past few years, as powerful AI chatbots disrupt how teachers teach and students learn.

His discipline – an emerging academic field called “public humanities,” which integrates cultural theories with a focus on community-building – is novel enough to require constant revision of its curriculum.

But the sheer speed at which artificial intelligence has evolved since November 2022 – when OpenAI released its powerful ChatGPT, kicking off an AI frenzy – has necessitated a more radical reimagining of higher education, Li said.

For example, with AI, writing essays has become less meaningful as a form of assessment, as the technology can generate prose “within seconds,” Li, a research assistant professor at the Chinese University of Hong Kong (CUHK), told HKFP in an interview in April.

Now, a course needs to incorporate several grading methods, including in-class handwritten tests, to fully assess students, he said.

The point is not to ban students from using AI, as Li himself allows students to freely use the technology in some settings.

“The logic is to make all these assignments interconnected and accumulative, so that [students] cannot instantly generate something” for the sake of handing in their homework, he said in Cantonese.

Across academic disciplines, professors and lecturers in Hong Kong are rethinking their teaching and the ways they assess students’ learning outcomes. The goal is to cultivate what scholars call “AI literacy” – a responsible and constructive approach to using the emerging technology in the real world.

“When [students] graduate, it is certain that they will be exposed to AI and they will use AI,” George Ho, an associate professor of supply chain and information management at the Hang Seng University of Hong Kong (HSU), told HKFP in Cantonese.

“What we need to do is to ensure they use it ethically, use it well, and acquire the necessary skills to master the technology,” Ho said.

The process is both fruitful and challenging, according to teaching academics and experts interviewed by HKFP. While generative AI has unleashed massive potential for innovative teaching and greatly accelerated students’ learning, the technology also raises questions about fairness and academic integrity.

From ban to embrace

Kelvin Wan, a digital learning specialist at HSU, recalled that when the AI boom began in late 2022, universities in Hong Kong were cautious about the technology and its implications for education.

That anxiety was demonstrated when the University of Hong Kong (HKU), the city’s oldest university, issued a temporary ban on AI tools for coursework in February 2023. HKU lifted the ban in the summer of that year and gave its teachers and students access to ChatGPT in September 2023.

Cecilia Chan, a professor of education at HKU, who leads the university’s AI in Education Lab, told media outlet The Initium in 2024 that the temporary AI ban was implemented to give school management time to acquire the hardware, software, and knowledge required to apply the nascent technology in education.

HKU has since fully embraced generative AI. In its AI policy, rolled out in the 2023-24 academic year, the university stated that “AI literacy is essential,” and that “rethinking pedagogy is most likely necessary.”

“Universities in Hong Kong were in fact watching each other at that time,” Wan told HKFP in Cantonese. “So we were pretty cautious at the beginning… But, from our perspective, we wanted to know what [AI] is.”

In May 2023, HSU organised a seminar on the pros and cons of using AI in higher education, as well as the ethical issues arising from the technology. Teachers were positive about the discussion, and the school began drafting guidelines for the use of AI for both staff and students, Wan recalled.

At HSU, the school encourages teachers to openly communicate with students about where they are allowed to use AI for assignments from the start of the semester, Wan said.

Meanwhile, the school also provides guidance for students to responsibly report their use of AI to teachers. “This hopefully will safeguard their academic integrity,” Wan said.

Ho, the supply chain professor at HSU, said his students were trained in skills to use AI, including systems thinking – the ability to break down a complex problem into solvable parts – and critical thinking – being able to evaluate the information generated by AI.

Wan called those “transferable skills,” meaning they are valid in various real-world scenarios, not just for using AI. “No matter how advanced AI becomes in the future, we as humans have to evaluate its outputs and make our own judgements,” he said.

Ultimately, students have to be responsible for their use of AI, Wan said. “AI is fundamentally a tool, and it cannot take responsibility… Whoever uses AI has the responsibility to criticise and evaluate its outputs.”

Subjects at risk

As AI takes root in classrooms, it appears to be a promising tool for teachers and students alike.

Tang Yuk-ming, a senior lecturer at the Department of Industrial and Systems Engineering at the Hong Kong Polytechnic University, has studied the impact of advanced information technologies on students’ learning outcomes and efficacy.

His research during the Covid-19 pandemic, when online learning was the predominant mode of higher education due to social distancing rules, found that virtual reality (VR) could improve students’ academic performance and motivation to learn.

AI chatbots can enable teachers and students to “personalise learning,” said Tang, whose research focus also includes the application of AI.

As the digital assistant is available 24 hours a day, students can get instant responses tailored to their needs. At the same time, teachers can review students’ interactions with AI to identify their weak spots for strategic intervention.

Tang’s department has developed in-house AI chatbots for students, and teachers will only use the data collected for teaching enhancement, he said, brushing off privacy concerns.

See also: AI as ‘personal therapist’: Despite risks, Hong Kong teenagers turn to chatbots for counselling

Ho shared the positive assessment of AI’s impact in his discipline of supply chain science. Thanks to AI’s assistance, what used to take a year for students to complete, such as problem-solving projects using real-world corporate data, can now be finished within a single semester.

But there has been a growing concern that, at the current rate of development, AI could shortly replace jobs and wipe out whole categories of work. STEM subjects – science, technology, engineering, and mathematics – are considered to be at risk.

Between 2022 and 2025, entry-level jobs in Hong Kong plummeted by 60 per cent, with junior roles in administration and programming vanishing at even higher rates, government data showed last month. The city’s labour minister, Chris Sun, said these areas are vulnerable to automation and that authorities are studying the impact of AI across job markets.

Both Tang and Ho dismissed the idea that AI would replace engineering graduates in the future and said that “tedious jobs” may disappear.

“My view is that [AI] will vastly raise productivity, but there may not be as many people with the skills to use these tools to finish tasks,” Tang said.

“STEM subjects are never just about training your basic knowledge; it is about your mindset and problem-solving abilities,” he added. “People who know how to solve problems will always be needed.”

For the humanities, the idea of AI as a threat is less acute, because human-centred knowledge and human interactions – such as engaging with a community – are some things AI cannot produce on its own, Li said.

“Ultimately, what we [in the humanities field] are interested in is people’s unique experience, how it can help us make sense of ourselves and what we can do as humans.”

In contrast, he said, AI can only assist students in this process.

Challenges in higher education

While AI shows much potential for improving teaching and learning, its adoption in higher education comes with its own challenges.

Wan said that, in his experience promoting AI literacy, teachers are often more adept at picking up the technology than students. “I realised our colleagues and teachers have to use [AI] every day at work, while students stop using it at semester breaks,” he said.

He also noted that university freshmen are typically less savvy at using AI for their learning compared with more senior students. The problem, he argued, stemmed from a lack of AI-related training in the city’s secondary school curriculum.

“DSE students are not that AI-ready coming out of secondary schools, so they may struggle when they enter universities and have to use the technology,” he said. “The university should provide some support in that regard.”

Another challenge lies with ensuring fairness as the use of AI becomes ubiquitous. Universities in Hong Kong have sought to provide students with equal access to cutting-edge AI models, but, as Li observed, a minority of students choose not to use the technology for their learning.

“It may be an ethical choice. They may think they have abused water resources, or they may think AI has made them dumb,” Li said of the students.

AI critics have raised concerns about how data centres powering the technology are using millions of litres of water for cooling, depleting the resource in some communities around the world.

There have been debates about AI’s impact on the mind. While some studies suggest that the tool can improve academic performance, other research has found that relying on AI for tasks like essay writing can weaken cognitive activities in the long run – an accumulated “debt” for the brain.

Both the costs and benefits may be real. According to a study published in April, while using AI improved people’s problem-solving in the short term, it impaired their independent performance and made them more likely to give up when not assisted by the technology.

The students who opt out of using AI “may see learning as a form of self-cultivation, and they don’t want AI to interfere with that process,” Li said. “It is an important reminder to respect this group of students when we design our courses.”

As AI grows more powerful at an exponential rate, there are also worries that teachers themselves are at risk of being replaced.

While this is an “interesting” possibility, Tang said, chatbots that are always available cannot “completely” replace teachers and schools.

“Teachers are not only about teaching. They understand the students’ needs through communication, and they guide students towards addressing those needs,” he said.

“AI cannot do this alone so far.”

More than 90 per cent of Hong Kong companies, schools, and NGOs have incorporated artificial intelligence (AI) into their workflows, according to a survey.

According to a survey of 800 organisations conducted by the Hong Kong Internet Registration Corporation Limited (HKIRC), 94 per cent said they had used AI tools.

Among those, 63 per cent had not established an internal AI usage policy for employees, while 68 per cent had not conducted any AI training, the survey found.

HKIRC CEO Wilson Wong said on Tuesday that employees at almost half of the surveyed organisations had used unauthorised AI tools.

“While the penetration rate of AI in the workplace is exceptionally high, most organisations still face security risks regarding governance, tool usage and training,” Wong was quoted as saying in a statement issued by the government’s Digital Policy Office (DPO).

He was speaking at a joint press conference on cybersecurity, alongside representatives from the DPO, the Hong Kong Police Force’s Cyber Security and Technology Crime Bureau, and the Hong Kong Computer Emergency Response Team Coordination Centre (HKCERT).

Security risks

Wong cited an earlier survey by the HKCERT, which found around 35 per cent of businesses using AI admitted to feeding company information into AI tools.

Some employees used open-source AI tools to process meeting minutes, for instance, which could lead to errors or data leaks, he added.

Wong said the HKIRC, which oversees Hong Kong domain names, would launch the Secure AI@Work Enablement Campaign to provide training and assistance in formulating AI policies, as well as suggestions for suitable AI tools and regulations on information that should not be processed by AI.

The campaign “will assist organisations in plugging governance gaps through training, AI strategy and policy formulation tools and advisory services,” the statement said.

Edmond Lai, chief digital officer of the Hong Kong Productivity Council, the parent organisation of the HKCERT, said that the HKCERT would seek to bolster public education and talent cultivation in AI and cybersecurity through publicity campaigns, such as AI-generated tram advertisements and videos.

Full-time job vacancies suitable for Hong Kong university graduates have plummeted by 60 per cent, as artificial intelligence (AI) sweeps through the city’s labour market, a minister has said.

Secretary for Labour and Welfare Chris Sun said on Wednesday that entry-level jobs vulnerable to automation have been hit hardest, with vacancies in administration dropping nearly 90 per cent over the three-year period and roles in information technology and programming falling by 80 per cent.

The number of full-time job vacancies suitable for university graduates shrank from 80,000 in 2022 to just 31,000 in 2025, the minister said.

The figures were derived from the Joint Institutions Job Information System, an online job search platform for students from Hong Kong’s eight publicly funded universities seeking employment, Sun said in his reply to enquiries by lawmaker Priscilla Leung.

“We all know the impact of AI is sweeping and global. We are all exploring how to help young people find jobs in a world changed by AI,” Sun told the Legislative Council in Cantonese.

Citing a survey by global consulting firm International Data Corporation, Sun said over 60 per cent of companies surveyed around the world had indicated they would cut entry-level positions in the next three years due to AI.

He vowed that the Labour and Welfare Bureau would analyse the impact of AI on Hong Kong’s overall labour market and specific industries.

Findings are expected to be released in the fourth quarter of this year as part of the mid-term update of the government’s Manpower Projections, he added.

He also said that, between 2025 and 2028, the eight University Grants Committee-funded universities will introduce 30 new academic programmes covering emerging sectors, such as AI, cybersecurity, and the creative industries.

Sun noted that, despite the drops in job vacancies, the unemployment rate among university graduates has not increased significantly.

The number of employed people aged 15 to 29 with a degree or above was about 268,000 in 2025, compared with 270,000 in the previous year, Sun said, citing government data.

Since OpenAI launched ChatGPT in November 2022, the use of generative AI chatbots and tools has become increasingly common across industries around the world.

Hong Kong Chief Executive John Lee has pushed for expanding the applications of AI across government departments and social sectors. In his Policy Address last year, he said the authorities would promote “extensive and deep integration of AI” across industries.

During his annual budget speech in February, finance chief Paul Chan announced that he would chair a new “AI+ and Industry Development Strategy” committee.

The government will also provide “AI training for all,” embedding AI education at different levels of education and vocational training, Chan said at the time.

By Emily Wang

Beijing cleaner Lin Meiqiong found her work a little easier the day she was paired with an unlikely new colleague — a tall, wheeled robot with AI-powered tidying skills.

The 56-year-old and her white-and-silver partner, fitted with cameras and two mechanical claws, are part of a new human-robot cleaning service offered by Chinese household help platform 58.com.

It’s a baby step towards a future espoused by tech evangelists in which robots increasingly take over manual labour from humans — though at the moment, such services are largely a data-gathering exercise for companies and a novelty for curious customers.

“It’s definitely different,” Lin told AFP in between cleaning the kitchen and wiping down windows.

“I used to have to do everything myself,” she said. “It’s reduced the workload a bit.”

The cleaning service, a collaboration between 58.com and Chinese robotics company X Square, costs 149 yuan (US$22) for three hours and is available in Beijing and tech hub Shenzhen.

Helped into the apartment by an X Square engineer, the AI-operated Quanta X1 Pro robot uses its cameras to identify areas it could spruce up.

As Lin scrubbed the floor on her knees, it picked up rubbish and folded clothes strewn across a sofa.

Grasping a pair of dark grey trousers, it raised its upper body to stretch the fabric taut, before laying it flat and arranging it into neat halves.

The process took several minutes and resembled a child learning to fold clothes for the first time.

Future iterations of the robot will respond to voice commands and even be able to chat, said the engineer, Hu Bowen.

‘Better than a lab’

Around 200 households have booked the service since it was rolled out in March.

Tan Pei, who works in advertising and booked the robot to clean her Beijing flat, said she had chosen the service because she was interested to “see what it could do”.

“Even though it’s not that perfect, there are still parts of it that surprised me,” such as folding a pair of trousers “quite well”, she said.

China’s robots have wowed audiences with fluid dancing and set-piece martial arts displays onstage, but their application and performance in real-life settings remains limited.

For companies like X Square, the logic of launching an imperfect service lies in data collection for so-called embodied artificial intelligence.

Unlike large language models trained on vast quantities of internet content, robots lack comparable real-world datasets.

“We don’t have a robot internet yet,” Christoforos Mavrogiannis from the University of Michigan told AFP.

“It is much more informative to put the robot out there and study what happens than staying forever in the lab.”

X Square engineer Hu said he sends his robots to work in a “completely unfamiliar environment”.

“That is very challenging, but this unfamiliar data is also very helpful for the robot’s growth.”

As investment into embodied AI booms, similar trials in China include robots directing traffic in cities like Hangzhou or working on factory floors.

On the domestic help front, firm GigaAI also plans to deploy 100 humanoid robots into households in central Wuhan this autumn for free home-service trials.

Investors have poured more than 57.7 billion yuan (US$8.5 billion) into China’s embodied AI industry so far this year, already soaring past the total for last year as a whole, according to business database ITjuzi.

‘Very elementary stage’

But a myriad of hurdles stand in the way of widespread deployment.

As the Quanta X1 Pro’s clothes folding demonstrated, robots still can’t match human dexterity.

“Even though many companies are working on building better hands and building autonomy for hands, we don’t have that yet,” the University of Michigan’s Mavrogiannis said.

There are multiple regulatory issues even once the physical capability is there.

Privacy will become a big issue, as robots would have access to huge amounts of personal data.

“We don’t know where that data is going, where it’s located… who is looking at that information,” said Valeria Alessandra Macalupu Chira from Queensland University of Technology.

The safety of clients and their homes is another unresolved issue.

“I think we are still at a very elementary stage,” said Yang Jianfei from Singapore’s Nanyang Technological University.

Robots currently require supervision by humans who can activate emergency stop functions, he noted, and there are not yet recognised industry-wide safety standards.

Experts agree broad adoption seems a long way off.

Asked whether she thought robots would revolutionise her industry, cleaner Lin did not seem too concerned.

“Compared with people, it’s obviously still not quite there,” she said. “After all, it’s a robot.”

Several times a day, a drone carrying high-capacity cameras and flashing red-and-blue lights whirs and rises from the rooftops of police stations across Hong Kong.

They emerge from a box-shaped docking system that slowly unfolds its doors to both sides. Some hover over the city’s billion-dollar villas with private pools and tennis courts; others whizz along streets bustling with people and traffic. 

Hong Kong police have been rapidly expanding their use of surveillance technology and automated drones. They have used drones to hunt down people who overstayed visas or gambled illegally.

According to the police force, these technologies will help deliver high-quality police services and optimise deployment and efficiency. Drones and cameras alike will also likely be combined with the use of artificial intelligence (AI) and facial recognition capabilities.

Drones will substitute some of the police’s foot patrols, and tens of thousands of surveillance cameras will be installed to assist in investigations and arrests. 

Since a drone patrol pilot scheme was rolled out in May last year, the technology has helped arrest 54 people, including at least six wanted individuals, according to police. The force did not provide complete data, but at least half of the suspects allegedly committed non-violent crimes.

In one operation last month that spanned from West Kowloon to Lantau Island, police – with the help of drones – arrested 19 people suspected of immigration infractions and prostitution. 

In another instance, police used drones to apprehend a group of eight middle-aged and elderly people who were gambling illegally in a public housing estate in Ma On Shan. Police also fined two drivers spotted crossing over into an oncoming traffic lane on a road to Shek O, using a drone. 

Police have not responded to HKFP’s request for more details on how the drones helped during those arrests and investigations. 

The increased use of drones is a response to China’s push for a “low-altitude economy,” which can be integrated into daily services ranging from deliveries to law enforcement, said Sky Yeung, chairperson of the DNT FPV Drone Association Hong Kong, China. 

Businesses such as delivery companies and government agencies can test drone-use scenarios through a regulatory exemption scheme, and the government is taking steps to prepare for more drones in the air, whether operated commercially or by authorities, Yeung said. 

So far, police have not explicitly said anything about using drones for national security purposes, which has been a priority for Hong Kong’s law enforcement in recent years.

However, as an expert told HKFP, the capability is there.

Despite the stated purpose of police technology, once the law allows for an agitator, a national security risk, or a terrorist to be prosecuted, it becomes “malleable,” said Bryce Neary, former executive editor of the Seattle Journal of Technology, Environmental & Innovation Law. The US-based lawyer studied the use of drone and surveillance technology in Hong Kong, China, and the US.

If a government “make[s] a legal argument to do so, then the technology is in place and can be utilised regardless, and as needed essentially, when the government wants to change those terms for their use,” Neary told HKFP on the phone.

Moreover, there are potential privacy issues.

To people on the ground, police drones flying between 60 and 90 metres above ground will be barely noticeable to the naked eye, Yeung said, and their buzzing noise is unlikely to cause a nuisance, given other urban noise. 

But nothing escapes the drones flying above us. Police drones, similar to those used in China, can typically “film everything” with “powerful lenses that can zoom in from a great distance, such as seeing what is inside a vehicle,” he said. 

Police drones are marked with flashing lights and reflective decals for people to identify them, but these won’t be visible at their usual operating altitude, Yeung said. “Maybe you can see a flashing dot at night, but you wouldn’t notice it.” 

When asked about privacy concerns by HKFP, the police force said its drone patrols fly over “carefully” planned routes that cover only public areas and do not involve private spaces such as building interiors.

The drones “avoid unnecessarily flying close to individuals or private premises” – unless the situation warrants an investigation, in which case the drones would descend to lower altitudes to collect evidence, police said in a statement.

Video footage with no evidential value will not be kept for more than 31 days, and those obtained as evidence will be classified as such and handled by the investigating unit, according to the police statement. 

Over the past two years, Hong Kong authorities have been introducing more surveillance technologies without hiccups at the “patriots only” legislature – and without protest.

This is in stark contrast to the time when angry demonstrators tore down experimental “smart” lampposts during the city’s 2019 protests and unrest. Discontent with shrinking political freedoms, protesters suspected that the lampposts would eventually allow authorities to conduct surveillance by adding facial recognition capabilities to their panoramic cameras. 

The government strongly denied such plans at the time, and promised the cameras would be disabled or their resolution reduced to assuage concerns. 

However, in a reversal, law enforcement is now considering adding facial recognition technology to its toolkit. Such systems may be connected to police surveillance cameras as soon as this year, police chief Joe Chow said in February. By 2028, police will install a total of 60,000 cameras across Hong Kong. 

The goal is to have “as many cameras as possible” and replicate what’s in mainland China, where there is camera coverage “every two steps,” he said during a TV interview. 

“Times have changed” compared with when society widely opposed increased surveillance and privacy issues, Chow added.

According to police, drone patrols will be used to combat crimes, identify traffic violations, and monitor traffic flows and crowds.

They can be used to track down suspicious individuals, such as someone who appears evasive when police are nearby, police said at press briefings.

They may also soon be equipped with artificial intelligence, but police have not specified whether the same facial recognition technology used on cameras would be applied to drones. 

Present technology from mainland Chinese drone surveillance vendors can identify people, objects, behaviours, and events, according to their product catalogues. They can count and identify various types of vehicles moving on a road, or people in a crowd. They can detect illegally parked cars, smoke, or objects fallen onto power lines. They can spot when protest banners are unfurled. 

Yeung pointed out drones’ ability to lock on to a target person and track them automatically as they move – a feature commonly used by police in the US. In short, drones film from above, while police operate on the ground. 

During the first phase, which began in May last year, drones were deployed to Heung Yuen Wai, a border area with mainland China, and West Kowloon.

In the second phase of the scheme, launched in January this year, police drone patrols were expanded to remote areas, where foot patrols are less frequent and which are more prone to burglaries, such as outlying islands like Lamma Island and Cheung Chau, as well as the Peak.

They also covered busy downtown districts like Central and suburban residential areas such as Yuen Long and Tsuen Wan.

The police force purchased around 700 drones for HK$25 million during the past financial year, and will purchase another 56 in 2026-27 for HK$4.8 million, the Security Bureau told the legislature.

Other agencies also deployed drones for various purposes, from detecting sites at risk of landslides to patrolling several tourist hotspots during Golden Week holidays.

Last year, investigators from the Food and Environmental Hygiene Department apprehended individuals who illegally slaughtered a goat in a rural area, with the help of drones.

Several residents on Lamma Island said they were not aware of police drone patrols, despite prominent banners announcing their presence near the Yung Shue Wan ferry pier and in villages. They said they welcomed the idea that these patrols could prevent bike theft or burglaries, and expressed no worry about privacy issues. 

The island’s resident, who asked to be identified only as Mark, said he believes ultimately it is the presence of police officers that will make a difference in deterring crime, something that drones above his head cannot replace. “What you need is your bobby to be walking and to be visible,” he said. 

Neary believes a chilling effect is the intended purpose of police drones, more than the number or severity of crimes they manage to actually solve.

“Regardless if it’s actually effective in terms of what it’s doing, the fear of the fact that you’re being monitored at all times for any of these petty crimes in public or in private, I think, is going to be a big deterrent for you to do so,” he said. “And maybe that’s the point in the first place, right?” 

A new UN study has named Hong Kong’s data centres as some of the most carbon-intensive in the world, blaming the city’s heavy dependence on a fossil-fuel-powered energy grid.

The report, titled “Environmental Cost of AI’s Energy Use,” examined the global carbon, land and water impacts of the infrastructure powering AI, saying that by 2030, data centres could consume 945 terawatt-hours.

That is “nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria, countries collectively home to more than 650 million people,” according to a UN press release.

“Indonesia, India, and Hong Kong (SAR) are among the most carbon-intensive grids with carbon footprints 62%, 51%, and 43% higher than the global average, respectively. Poland and Mainland China rank lower with carbon intensities at 30% and 21% higher than the global average,” the UN University’s Institute for Water, Environment and Health said in a report on Wednesday.

In comparison, the carbon footprint of electricity in the US, Germany, and Italy is 18 per cent, 24 per cent, and 32 per cent below the global average, respectively.

Energy in Hong Kong is 67 per cent derived from fossil fuels, 32 per cent from nuclear and just 1 per cent from renewables, the report said.

There is also a water footprint for cooling heat-intensive data centres, as well as a land footprint. “AI-related water consumption could equal the basic annual domestic needs of 1.3 billion people by the end of the decade, while its land footprint may exceed 14,500 square kilometres – roughly twice the size of the Jakarta metropolitan area,” the UN said.

However, Hong Kong was ranked among the lightest for water and land consumption, mostly because its energy mix does not rely on renewable energy sources, which require large amounts land.

As a trade and logistics hub, with around 300 internet service providers, Hong Kong remains a prime location for data centres. Its telecommunication networks connect to 12 external submarine optical fibre cable systems, with more under construction, according to the city’s Digital Policy Office.

The government is building a new 110,00 square metre data facility in Sandy Ridge, 90 per cent of which will be dedicated to data centres, according to a government press release in March.

Daily AI use, not training

The UN report said that day-to-day use of AI models accounted for around 80 to 90 per cent of total energy demand, as opposed to just model training. It cited the case of ChatGPT, which was processing around 2.5 billion prompts per day, with image generation requiring a thousand times more energy than a simple text query.

“China’s DeepSeek, launched in January 2025, attracted more than 20 million daily active users within three weeks, and had about 125 million monthly active users by mid-2025,” the report said.

According to the Digital Policy Office website, “data centre operators are all striving to enhance energy efficiency , so as to reduce their power consumption, their operating expense and also their environmental impacts.”

It cites existing policies by the Electrical and Mechanical Services Department (EMSD), which set rules for ensuring the energy efficiency of buildings and regulate the use of fresh water in cooling towers for air conditioning systems.

The EMSD’s Green Data Centres Practice Guide lays out initiatives for efficient data centre design, procurement, operations and disposal, whilst also promoting the use of assessment tools to measure environmental impacts.

The United States issued an updated list on Monday of Chinese companies that it believes are aiding the country’s military — including e-commerce giant Alibaba, search engine provider Baidu and electric vehicle maker BYD.

The US Defense Department unveiled the designations just weeks after President Donald Trump met Chinese leader Xi Jinping in Beijing, with both sides seeking to maintain stability in the bilateral relationship.

Trump has since invited Xi to pay a reciprocal visit to Washington in September.

But the latest release could fan tensions between the world’s two biggest economies.

The latest Pentagon update came months after it briefly released — then withdrew — an earlier version of the list without explanation.

The new list is largely similar to the version momentarily published in February, although two memory chipmakers were reinstated to the blacklist after having been removed from it at the time.

The re-added companies are ChangXin Memory Technologies and Yangtze Memory Technologies.

“This updated list of Chinese military companies is a warning to American businesses, all levels of government, and the American people,” said Representative John Moolenaar, the Republican chair of the House Select Committee on China.

He urged in a statement for US companies to “stop doing business with these threats to our national security” or risk “enabling China’s military ascendance.”

The companies targeted also cover some of China’s key tech giants involved in artificial intelligence, including Alibaba, Baidu and Tencent. Tencent was already previously designated.

Baidu opposed the list in a statement on Chinese social media, calling the accusations “entirely baseless”.

“We categorically reject the inclusion of Baidu on the list, and there is no credible justification for adding Baidu to the list,” a spokesperson said.

“The suggestion that Baidu is a military company is entirely baseless. We will not hesitate to use all options available to us to have the company removed from the list.”

Alibaba called its inclusion in the list “a mistake”, threatening legal action.

“There is no basis to conclude that Alibaba Group should be placed on the CMC List. Alibaba Group is not a Chinese military company nor part of any military-civil fusion strategy,” the company said in a statement.

While the determinations have few immediate legal implications for many of the companies, it is seen as a move that could precede more punitive measures.

Other companies that were added include pharmaceutical firm WuXi AppTec and start-up Unitree, which makes humanoid robots.