News related to
developers

The stock climbed about 6 per cent on Tuesday as investors reacted to signs that Tencent is moving closer to embedding a task-performing assistant inside WeChat, the messaging, payments and social platform used by more than 1.4 billion people. The prototype is designed to help users complete everyday actions within the app, including finding venues, interacting with mini-programmes and placing customised orders, rather than merely answering questions in chatbot form.

The planned agent would mark a significant step in Tencent’s push to convert its AI investments into consumer products at scale. WeChat already functions as a digital operating system for daily life in China, covering messaging, payments, shopping, food delivery, public services, entertainment, ride-hailing and enterprise communication. That reach gives Tencent an advantage that few rivals can match: a single platform with a vast user base, heavy daily engagement and a dense network of third-party services.

People familiar with the prototype have indicated that users would be able to access the AI agent through a swipe gesture on the main WeChat screen. Tencent is expected to begin the regulatory compliance process before wider testing, with a small group of external users likely to receive access before any phased public launch. A firm release date has not been set, reflecting both regulatory uncertainty and the technical demands of deploying agentic AI across a platform of WeChat’s size.

The market response underscores how investors are reassessing Tencent’s AI position after months of concern that the company was trailing Alibaba, ByteDance and Baidu in the race to commercialise generative AI. Tencent has promoted Yuanbao, its AI chatbot, developed its Hunyuan foundation model and introduced productivity-focused AI tools, but WeChat remains the asset most capable of turning those efforts into a daily consumer habit.

Tencent’s first-quarter earnings showed both the strength and the cost of that strategy. Revenue rose 9 per cent year on year to RMB196.5 billion, while IFRS net profit attributable to equity holders increased 21 per cent to RMB58.1 billion. Marketing services revenue rose 20 per cent to RMB38.2 billion, helped by AI-driven advertising improvements, while domestic games revenue grew 6 per cent and international games revenue climbed 13 per cent. Capital expenditure reached RMB31.9 billion, up 16 per cent, as spending on AI infrastructure continued to rise.

The company has told investors that new AI products are weighing on margins, even as core businesses continue to generate strong cash flow. Excluding new AI products, non-IFRS operating profit rose 17 per cent in the first quarter, compared with a 9 per cent increase overall. That gap highlights the near-term drag from AI investment, but also explains why investors are sensitive to any sign that Tencent can place AI directly into revenue-generating services.

Competition is intensifying across China’s technology sector. Alibaba has pushed Qwen into consumer and enterprise markets, ByteDance has expanded AI features across its content ecosystem, and Baidu continues to build around Ernie and AI cloud services. The key battleground is shifting from chatbots to agents that can perform tasks, connect with apps and act across digital services with limited user input.

WeChat’s mini-programme ecosystem could be central to Tencent’s approach. Millions of merchants, developers and service providers already operate within the platform, allowing an AI agent to navigate services without pushing users outside Tencent’s environment. For merchants, an effective agent could increase conversion rates by shortening the path from search to purchase. For Tencent, it could deepen engagement, improve advertising relevance and strengthen its fintech and business services ecosystem.

Regulation remains a critical constraint. China requires generative AI services offered to the public to comply with rules covering data security, content control and algorithmic governance. Any WeChat agent would need to satisfy those requirements before broad release, especially because it could interact with payments, commerce and public services.

The Provisions on the Protection of Trade Secrets, issued by the State Administration for Market Regulation under Order No. 126, came into force on 1 June 2026 and replace rules first introduced in 1995. The overhaul gives regulators clearer authority to handle disputes involving digital assets that now sit at the centre of corporate value, from AI model architecture and training datasets to platform recommendation systems, software code and commercially valuable operational data.

The change marks a significant shift in China’s treatment of intangible technology assets. Earlier rules were drafted before cloud computing, AI systems and large-scale data operations became central to business competition. The new framework defines trade secrets as technical or business information that is not publicly known, has commercial value and has been subject to confidentiality measures by the rights holder. Technical information now expressly covers data, algorithms, computer programs and code, while business information can include management, sales, finance, planning, customer and operational data.

The timing gives the regulation wider strategic weight. China and the United States are locked in a prolonged technology rivalry involving semiconductor controls, restrictions on advanced computing hardware, limits on investment in sensitive sectors and scrutiny of cross-border technology transfers. Washington has tightened export controls on high-end chips and chipmaking equipment, while Beijing has expanded its own legal toolkit to protect critical technology, secure data flows and prevent overseas transfer of capabilities it regards as strategically important.

The new rules do not create a separate trade secret law, but provide more detailed administrative guidance under China’s Anti-Unfair Competition Law. They clarify how regulators should assess whether information is secret, whether it has commercial value, what counts as reasonable confidentiality measures and how infringement should be investigated. For companies, the central message is that protection will depend not only on the value of the information, but also on whether internal controls, access limits, confidentiality agreements, digital safeguards and evidence trails are in place.

The provisions also recognise that unsuccessful research can have commercial value. Stage-by-stage technical results, failed experimental data and development paths that save time or cost in future research may qualify for protection if they meet the legal conditions. That is especially relevant for AI labs, pharmaceutical developers, electric vehicle makers, chip design firms and advanced manufacturing groups, where failed trials can reveal important technical boundaries and prevent competitors from repeating expensive mistakes.

Regulators have been given practical enforcement tools, including powers to inspect business premises, question individuals, copy documents, examine electronic records and take action against materials linked to suspected infringement. Penalties can include orders to stop unlawful conduct, confiscation of illegal gains and fines that may reach CNY5 million in serious cases. The framework also covers improper acquisition, disclosure, use or permission to use trade secrets obtained through theft, bribery, fraud, coercion, electronic intrusion or breach of confidentiality obligations.

For multinational companies operating in China, the new rules create a mixed picture. Stronger protection could help firms defend proprietary data, software systems and industrial processes in a market where trade secret disputes have often been difficult to pursue. At the same time, wider protection for data and algorithms may increase compliance complexity, particularly in joint ventures, outsourced research, cloud services, employee mobility, cross-border collaboration and licensing arrangements.

Technology companies are likely to face tougher expectations over documentation. Firms seeking protection will need to show that the information was not easily accessible, carried economic value and was actively protected. Generic claims that an algorithm or dataset is confidential may not be enough. Companies will be expected to classify sensitive assets, restrict access by role, track downloads and transfers, maintain logs, use encryption or watermarking where appropriate, and ensure that employment and partnership contracts define confidentiality obligations with precision.

The rules also sit alongside China’s broader data and security regime, including laws on data security, personal information protection, cybersecurity and export control. That overlap could make enforcement more powerful but also more complicated. A dataset may be a trade secret, a regulated data asset and a cross-border transfer concern at the same time. Businesses handling AI training material, user behaviour data or industrial performance data will need to assess those obligations together rather than treating trade secret protection as a standalone issue.

The Mach33: Physical AI Fashion Show, held on 28 May at Galaxy Robot Park in Gangdong District, was presented as more than a novelty catwalk. Galaxy Corporation framed the event as an experiment in coexistence, asking how people and robots might share cultural spaces rather than treating humanoids only as factory tools, laboratory platforms or service machines.

Each model appeared with a shorter humanoid counterpart, often wearing matching or complementary outfits. A tasselled blue Texan-style look, complete with a robot-sized cowboy hat, stood alongside retro silver outerwear, silky dresses fitted to rigid frames and space-age black trousers reminiscent of David Bowie’s 1970s stage style. The styling made the mechanical bodies appear less like exposed prototypes and more like performers entering a commercial entertainment format.

Galaxy Corporation founder and chief executive Choi Yong-ho said the company had realised that robots “need to wear clothes” and argued that robots, like people, should have distinct identities. The company designed the outfits and intends to launch them under the MACH 33 name by the end of the year, positioning robot fashion as a potential extension of its entertainment-technology business.

The event followed the opening phase of Galaxy Robot Park, a 16,500-square-metre venue in eastern Seoul designed around robot concerts, interactive shows, robotic portrait drawing, boxing-style demonstrations and visitor experiences. The park is part of Galaxy Corporation’s effort to create a new live-entertainment category built around “physical AI”, a term used to describe artificial intelligence moving from screens and software into embodied machines operating in shared human spaces.

Galaxy Corporation, known for managing cultural figures including G-Dragon, Taemin and actor Song Kang-ho, has been seeking to merge K-pop spectacle with robotics. Its robot performances have included choreographed dance routines, synchronised stage movement and plans for several shows a day. The company has said it wants to stage more than 1,000 robot performances annually and ultimately take robot-led shows beyond South Korea.

The robot models appeared to be Unitree G1 humanoids, a relatively low-cost platform made by Hangzhou-based Unitree Robotics. The G1 is widely used in demonstrations because of its compact size, balance control, range of motion and comparatively accessible pricing for developers and institutions. Its ability to walk, turn, gesture and respond to programmed choreography makes it suitable for controlled performances, though it remains far from a fully autonomous general-purpose worker.

The show also highlighted the gap between polished presentation and engineering reality. Humanoid robots are becoming more agile, but smooth movement in a runway setting still requires careful staging, reliable control systems and predictable conditions. Research into whole-body teleoperation, motion retargeting and autonomous loco-manipulation continues to point to unresolved challenges, including latency, balance recovery, hand dexterity, human-robot interaction and safe operation in crowded public spaces.

For South Korea’s entertainment sector, the Galaxy event fits a broader pattern of experimentation with virtual idols, AI-generated content, immersive fan platforms and digitally enhanced performances. K-pop’s visual emphasis makes it a natural testing ground for robotic spectacle, but the commercial test will be whether audiences view robot performers as compelling cultural products or as temporary curiosities.

Fashion offers Galaxy a softer route into that debate. Unlike industrial robotics, a runway show does not require humanoids to prove productivity. It asks whether machines can be styled, branded and given a visible role in popular culture. The clothes also conceal some of the awkwardness of exposed mechanical frames, making the robots appear less clinical and more theatrical.

The commercial stakes are growing. Forecasts for humanoid robotics vary widely, but major financial institutions and technology analysts expect the sector to expand sharply over the next two decades as costs fall and AI systems improve. Optimistic projections point to hundreds of millions, or even more than a billion, humanoids by 2050, though adoption is expected to move gradually as companies work through safety, reliability, regulation and public acceptance.