分类： technology

In a historic move that marks a watershed moment for global tech labor organizing, hundreds of Google DeepMind employees based in the United Kingdom have overwhelmingly voted to form a union, driven by urgent ethical concerns over the misuse of the company’s artificial intelligence technology by the U.S. and Israeli militaries in conflicts involving Iran and Gaza.

Following the internal ballot held among Communication Workers Union (CWU) members at DeepMind, official results showed a 98% majority in support of unionization. On Tuesday, workers formally submitted a request to Google DeepMind’s leadership to recognize both the CWU and Unite the Union as their official workplace representatives. If recognized, this organizing effort will become the first formal union at a major global frontier artificial intelligence research lab, according to campaign organizers.

At the core of the workers’ demands is an immediate end to the provision of Google AI tools to the Israeli military and U.S. defense entities. Additional demands include the reinstatement of a previously discarded company pledge to refrain from developing AI-powered weapons and mass surveillance tools, the establishment of an independent ethics oversight body with decision-making authority, and formal guarantees that individual employees retain the right to decline work on specific projects on moral or ethical grounds.

This unionization push is part of a broader global grassroots movement among DeepMind staff, with employees across international locations planning coordinated in-person protests and so-called “research strikes,” a work stoppage action where researchers suspend their regular work to highlight their concerns.

One anonymous DeepMind employee emphasized that no staff want their cutting-edge AI research to become complicit in violations of international law, noting that the technology is already aiding what the employee called Israel’s genocide of Palestinians in Gaza. “Even if our work is only used for administrative purposes, as leadership has repeatedly told us, it is still helping make genocide cheaper, faster, and more efficient,” the worker said. “That must end immediately, as must harm to Iranians and human lives anywhere.”

The successful unionization vote would cover at least 1,000 DeepMind employees based at the company’s London headquarters. In the formal letter submitted to management, workers have given DeepMind leadership 10 working days to voluntarily recognize the two unions or enter into mediated negotiation talks. If management fails to meet the deadline, workers will launch a formal legal process to force official recognition.

The vote comes at a time when the UK government is actively pursuing policies to attract major global technology investment, positioning the country as a leading hub for AI innovation. Already, several high-profile AI firms including Anthropic have announced plans to expand their London-based operations in recent months.

CWU national officer John Chadfield framed Tuesday’s announcement as a defining moment for tech workers around the world. “This is a really important moment where tech workers at Google’s frontier AI lab are connecting with some of the most oppressed people in communities around the world in meaningful ways, based on foundational values of solidarity and trade unionism,” Chadfield said. “By exercising their rights to collectivise they are in a strong position to demand their employer stop circling the ethical drain of military-industrial contracts, echoing the sentiment of many working people in the UK and elsewhere.”

This latest labor action is the culmination of years of growing internal unrest among Google employees over the company’s military contracts. Earlier this year, more than 600 Google staff across the company’s AI and cloud divisions signed an open letter urging leadership to cut off the Pentagon’s access to Google technology for classified military operations. Google previously drew widespread backlash after firing dozens of employees who participated in protests against Project Nimbus, a joint initiative with Amazon to provide cloud and AI services to the Israeli government amid the ongoing military campaign in Gaza. In 2023, the company also faced heavy criticism for its acquisition of Israeli cloud security firm Wiz, which was founded by former veterans of Unit 8200, the Israeli military’s elite cyber espionage and surveillance unit.

Against a backdrop of escalating wildfire risk driven by climate change, artificial intelligence is rapidly becoming a critical new tool for early wildfire detection across fire-prone regions of the western United States, already proving its ability to stop blazes before they turn into catastrophic infernos.

The technology delivered a convincing proof of concept on a March afternoon in Arizona’s Coconino National Forest, when an AI-enabled monitoring camera picked up a faint plume of smoke that did not match the signature of cloud cover or wind-blown dust. After human analysts confirmed the anomaly, alerts were immediately dispatched to Arizona’s state forest service and Arizona Public Service (APS), the state’s largest electric utility. First responders arrived on scene quickly and contained the resulting Diamond Fire to just 7 acres (2.8 hectares), a fraction of the size it could have reached if detected hours later.

The Diamond Fire interception is far from an isolated success. As record-breaking high temperatures and record-low winter snowpack stoke fears of an extreme wildfire season, state agencies, power utilities and private tech firms have been rolling out AI monitoring systems across remote, high-risk regions where human spotters or casual 911 reports often fail to catch blazes in their earliest, most controllable stages.

To date, APS already operates roughly 40 active AI smoke-detection cameras across Arizona, with expansion plans to bring the total to 71 by the end of the current summer. The Arizona Department of Forestry and Fire Management has also deployed seven of its own AI units, while Colorado-based utility Xcel Energy has installed 126 AI cameras, with goals to bring the system to all but one of the eight states it serves by the end of the year. In California, the ALERTCalifornia network operates more than 1,200 AI-integrated cameras that follow a detection model similar to Arizona’s system.

“Earlier detection means we can launch aircraft and personnel to it and keep those fires as small as we can,” explained John Truett, fire management officer for the Arizona Department of Forestry and Fire Management.

Unlike populated areas where residents often spot and report fires quickly, most high-risk unpopulated rural and remote zones lack consistent human observation. It is exactly these gap areas that AI monitoring fills, providing 24/7 surveillance that frequently outpaces 911 reports by a significant margin. According to Neal Driscoll, a geology and geophysics professor at the University of California, San Diego and founder of ALERTCalifornia, human oversight paired with ongoing AI training has kept false positive rates very low, and the technology consistently outperforms traditional 911-based detection timelines.

“It’s just the ones where we won’t get a 911 call for a long time, it is extremely helpful to have that AI always monitoring that camera,” said Brent Pascua, a battalion chief for the California Department of Forestry and Fire Protection (Cal Fire). “In many cases, we’ve started a response before 911 was even called, and in a few cases, we’ve actually started a response, went there, put the fire out, and never received a 911 call.”

Pano AI, one of the leading providers of this technology, integrates high-definition camera feeds, satellite data and artificial intelligence to scan for early smoke signs. Since launching in 2020, the company has seen surging demand for its systems, which are now deployed across 17 U.S. states plus Canada and Australia, serving forestry operators, public agencies and utilities including APS. In 2025 alone, the company’s technology detected 725 wildfires across the U.S.

“In many of these situations, we hear from stakeholders that the visual intelligence, the time, really, really gives them a head start and some of these could have taken off into hundreds if not thousands of acres,” said Arvind Satyam, Pano AI co-founder and chief commercial officer. APS meteorologist Cindy Kobold confirms the technology delivers an average 45-minute head start over the first incoming 911 call, a gap that can make the difference between a contained small blaze and a destructive megafire.

Satyam notes that the development of this technology was directly driven by the growing wildfire crisis fueled by climate change. Rising global temperatures from fossil fuel emissions have created drier, more fire-prone conditions that increase the frequency, intensity and speed of wildfire spread, and existing management tools have not kept pace. AI detection fills this gap, helping first responders act more effectively while protecting communities and critical infrastructure.

Despite its clear benefits, the technology still faces notable limitations and challenges. The most significant barrier for many agencies is upfront and ongoing cost: Pano AI charges approximately $50,000 per camera annually, a price that includes 24/7 monitoring support and fire risk analysis. False alarms remain a persistent issue, wasting valuable first responder time and resources even as training reduces their frequency. Even when the AI correctly identifies a fire, it cannot guide decision-making on response strategy – questions of when to deploy crews, whether to order evacuations, or how to prioritize resources still require human judgment and decision support systems.

In dense urban areas, where residents already report fires quickly, the technology offers less benefit, and it cannot keep pace with rapidly shifting fire behavior during extreme weather events such as the 2025 Los Angeles wildfires, when hurricane-force winds reshaped fire lines hourly. Proponents emphasize that AI is designed to complement, not replace, human firefighting teams.

“As the fire moves and shifts around, that’s where the human factor comes in and decides which tactics are best in fighting the fire. AI can only do so much,” Pascua said. “It just provides that real time information where we can make better decisions on the fire ground.”

AI’s role in wildfire management extends far beyond early detection, researchers note. The technology can already map optimal zones for controlled burns and vegetation thinning, and monitor air quality for early smoke signatures with far greater sensitivity than traditional consumer detectors. At George Mason University, professor Chaowei “Phil” Yang is leading a collaboration with California State University Los Angeles, the city of Los Angeles and NASA’s Jet Propulsion Laboratory to develop an AI system that forecasts wildfire spread and predicts which communities will face the worst smoke pollution impacts. The system will generate real-time maps to help agencies make faster, more effective decisions around evacuations, road and school closures, and public health warnings, with a target operational launch within three years.

Experts agree that AI integration in wildfire management is no longer a future concept – it is already deployed in active response, and its use will only expand in coming years. “AI in wildfires, it’s no longer just speculative. It’s really being used,” said Patrick Roberts, a senior RAND Corporation researcher who recently completed a study on innovation in wildfire management. “The future is AI everywhere, and the lines will blur between AI wildfire detection and just wildfire detection as the lines will blur in other areas of our life.”

China has formally blocked Meta’s proposed $2 billion acquisition of Manus, a high-profile Chinese general-purpose agentic AI startup, and moved to clear up misperceptions around the decision, emphasizing that the prohibition targets regulatory circumvasion rather than domestic firms’ legitimate overseas expansion or foreign inbound investment.

The ban was issued on Monday by the Office of the Working Mechanism for Security Review of Foreign Investment under the National Development and Reform Commission (NDRC), which ordered the involved parties to unwind the unreported transaction entirely. In the days following the ruling, Chinese state media outlets published a series of explanatory commentaries to outline the policy logic behind the decision, aiming to avoid misinterpretation that the move signals a broader crackdown on foreign capital or restrictions on Chinese tech firms going global.

Chinese policy analysts stress that Beijing does not intend for the Manus ruling to send a misleading signal to the global investment community. As a CCTV-affiliated social media account Yuyuan Tantian clarified in a Thursday article, China’s existing Measures for the Security Review of Foreign Investment draw clear boundaries for regulatory scrutiny. Under the framework, all investments touching on national defense security require mandatory declaration regardless of foreign stake size, while for key sectors including core information technology, internet products and services, and critical technologies, any transaction that grants actual control to a foreign investor falls within mandatory review scope.

Manus, the article noted, fits clearly into this defined key technology category as a developer of general-purpose AI agent systems. Meta’s proposed acquisition would have transferred full actual control of the startup to the US tech giant, yet neither party submitted the required proactive declaration to Chinese regulators, making the ruling a straightforward application of existing law.

The article added that Chinese regulators assess risk across three core dimensions: technology, talent and data. All of Manus’s core assets — including its foundational algorithms, training data and core R&D team — were developed by domestic teams within China’s borders, so any transfer of control overseas legally requires a national security review. The commentary also pointed to growing global trends of expanding security review scopes and blurred threat definitions that specifically target other countries’ AI development, a practice that China must guard against to protect its own strategic technology ecosystem. Even as it enforces security rules, China remains committed to supporting AI innovation and maintaining an open market for foreign investment, the article emphasized.

The Manus transaction grew out of a new regulatory workaround that has emerged since the United States barred American investment from China’s domestic AI sector in October 2024, dubbed “Singapore washing.” The term describes the practice of Chinese AI firms spinning off operations or relocating their registered headquarters to Singapore to avoid US investment restrictions and raise foreign capital. In the case of Manus, the startup restructured its operations to sever formal ties with its Chinese origins to secure Meta’s investment, a strategy that the ruling has now invalidated.

Manus first captured global tech industry attention when it made its high-profile debut in March 2025. Unlike conventional large language models such as ChatGPT or DeepSeek, Manus is positioned as a general-purpose AI agent capable of completing complex, multi-step tasks traditionally handled by white-collar workers. In promotional demonstrations, co-founder Xiao Hong showcased the system’s capacity to sort through 10 candidate resumes, identify a New York City property matching a set budget, and analyze stock correlation trends between Nvidia, Marvell Technology and TSMC, leading the startup to adopt the slogan “Leave it to Manus.”

The acquisition deal began taking shape in 2025, as Manus restructured to move its registered headquarters to Singapore between June and July that year. It reorganized under a new Singapore-based operating entity, Butterfly Effect Pte, reduced its mainland Chinese team from more than 120 employees to just 40 core members who were relocated to Singapore, deleted all Chinese-language social media accounts, and blocked IP addresses based in China from accessing its official website. By the end of 2025, Manus presented itself as a fully Singapore-based company, and Meta announced the $2 billion acquisition on December 30, with Xiao Hong slated to take a senior leadership role at the US firm.

Chinese regulators launched their formal review of the unreported transaction in January 2026, and by late March, Xiao Hong and co-founder Ji Yichao were barred from leaving China as the review progressed. The formal ban on the deal was issued on April 27.

According to Chinese analysts, Manus crossed three non-negotiable red lines in its restructuring and dealmaking: technology sovereignty, data sovereignty and national security. “Where the technology originates determines jurisdiction,” explained Guangdong-based business columnist Shengchandui. Manus’s core algorithms and core team were built entirely within China, so shifting the company offshore and selling it to a foreign buyer amounts to unauthorized export of domestically developed strategic capabilities, a form of “technology smuggling” that weakens China’s domestic innovation base. The columnist added that Manus processes vast volumes of user data, much of it originating from Chinese users, so transferring control overseas creates unacceptable risks of data leakage, particularly under existing rules governing cross-border data transfers. As AI agents are emerging as core infrastructure for digital work, communication and software development, putting a system built on Chinese technology and data under full foreign control creates unacceptable national security risks, he noted.

Zhu Youping, a researcher at the NDRC’s State Information Center, clarified that the ruling is not a restriction on legitimate global expansion by Chinese firms, but a prohibition on efforts to evade national regulation. “If the proposed acquisition is completed, Meta would obtain 100% control in Manus, but neither Meta nor Manus had declared this to the Chinese regulators,” he said. Regulators apply a “look-through” approach that focuses on the actual origin of technology, the source of training data and ownership of core talent, rather than just the jurisdiction where a company is registered. “Manus’s relocation to Singapore is essentially a case of using domestic resources to incubate value and monetizing it through an offshore structure to bypass oversight,” Zhu added.

Beyond blocking the unauthorized transaction, Chinese authorities have signaled that they want Manus to remain rooted in China to contribute to the country’s fast-growing domestic AI industry. In a Tuesday editorial, the Global Times noted that “China’s AI industry has entered a phase of rapid development, with a sustained burst of innovative vitality, making it a fertile ground for global AI innovation. We hope that more technology and innovation enterprises, including Manus, can find their place in this blue ocean in China, develop confidently, grow larger and stronger and achieve better development and breakthroughs.”

The Manus ruling aligns with Beijing’s latest policy push to scale up domestic AI adoption for economic growth. On April 21, China’s State Council released a policy document outlining 20 measures to expand and upgrade the country’s AI sector, setting a target of growing total industry output to more than 100 trillion yuan (approximately $13.8 trillion) by 2030, up from 81 trillion yuan in 2025. The policy specifically supports deployment of AI tools in high-impact areas including intelligent programming, contract review, financial services and supply chain optimization, and calls for the construction of national AI application testing bases.

Pang Chaoran, a researcher at the Chinese Academy of International Trade and Economic Cooperation (CAITEC), said the new policy marks a clear shift in China’s AI strategy: instead of focusing primarily on subsidizing AI model training, Beijing is now encouraging private service sector firms to adopt AI models and agents at scale. By driving widespread adoption of AI tools across industries, the government aims to accelerate commercialization of AI innovation, embed the technology deeper into real economic activity, and generate new growth momentum for both the service and technology sectors.

JAKARTA, Indonesia – In a move that marks one of the strictest youth safety policies the global gaming platform has ever enacted, Roblox confirmed Thursday it will require mandatory facial scanning for all Indonesian users under the age of 16 to verify their age, a change implemented to comply with Indonesia’s sweeping new regulatory framework governing minor access to social media and digital services.

Roblox Vice President and Global Head of Public Policy Nicky Jackson Colaco unveiled the new requirements during a Jakarta press conference, noting that the tailored rules for the Indonesian market outpace most other age-verification policies the platform has rolled out across its global operations. To align with national regulations, the company has restructured its Indonesian user accounts into two age-specific tiers: Roblox Kids, designed for children aged 5 to 12, which removes all in-platform chat functionality entirely; and Roblox Select, for teens aged 13 to 15, which restricts chat interactions exclusively to connections pre-approved by parents or family members.

The rollout will automatically reclassify the platform’s 23 million existing Indonesian accounts that were self-identified as belonging to users under 16, requiring all of these accounts to complete facial scanning-based age verification to retain their current access settings. Any under-16 user that fails to complete the facial scan process will be automatically downgraded to a restricted Roblox Kids account, with all chat functionality permanently disabled until verification is completed.

The age verification process requires users to capture a short video selfie to generate an estimated age assessment. Jackson Colaco emphasized that all biometric data collected during the process is deleted immediately after verification is complete, with no user data stored on Roblox servers long-term. According to Indonesian Communication and Digital Affairs Minister Meutya Hafid, Roblox has a total user base of roughly 45 million people in Indonesia, with just over half – around 23 million users – falling under the 16-year age threshold.

Notably, Roblox is the only gaming platform classified as a “high-risk” service by the Indonesian government, requiring it to implement more stringent youth access restrictions than most other major social media platforms operating in the country. Beyond account classification and restricted interaction limits, Roblox will also sort its game library by age appropriateness and enforce mandatory screen time limits to address widespread public concerns over youth gaming addiction. Parents will also be able to set custom daily usage caps aligned with their household rules, Hafid added.

Indonesia’s new national regulation on minor digital access took effect in late March, banning all users under 16 from accessing high-risk digital platforms that may expose young people to harms including gaming addiction, explicit content, online fraud, and cyberbullying. Out of eight major high-risk platforms operating in the country – which include YouTube, TikTok, Facebook, Instagram, Threads, X, and Bigo Live – seven have already committed to rolling out compliant age-based access restrictions. Alongside access limits, Indonesian regulators are pushing all digital platforms to publish regular disclosures of how many under-16 accounts have been restricted or suspended as part of the new policy’s implementation.

Starting this Friday, May 1, sweeping new drone regulations will go into effect across Beijing, introducing some of the strictest controls on unmanned aerial vehicles (UAVs) in the world, even as China positions the broader low-altitude economy as a key national strategic growth sector.

Under the newly implemented rules, the sale, rental, and unauthorised transport of drones and their core components into Beijing will be fully prohibited. All private and commercial drone owners operating within the capital are required to register their devices with local law enforcement before taking any outdoor flight. Additional requirements mandate that all outdoor drone operations across Beijing secure prior official approval, and operators must complete a government-designed online training course and pass an assessment on UAV regulatory policies to qualify for flight permits.

Cross-city movement of drones also faces new scrutiny: any drone brought into or out of Beijing must go through formal registration, and owners sending drones outside the capital for repair are required to collect the device in person after servicing, rather than accepting courier delivery. A narrow set of exceptions applies to officially approved public utility operations, including counter-terrorism missions and disaster relief response, where drone ownership and operation remain permitted following government authorisation.

This latest round of restrictions builds on years of incremental tightening of UAV rules across China, a trend driven by consistent official concerns over public safety and low-altitude airspace security. Beijing first designated its entire airspace a controlled no-drone zone last year, requiring advance air traffic management approval for all drone flights, and the local People’s Congress approved the updated regulatory framework in March 2026.

“Our goal is to strike the best balance between safeguarding public and airspace safety and supporting sustainable technological and economic progress,” explained Xiong Jinghua, a senior official with the Beijing Municipal People’s Congress, when the regulations were approved.

The new policy comes at a paradoxical moment for China’s drone sector: Chinese manufacturers, led by global industry leader DJI, currently dominate the worldwide consumer and commercial drone market, and the country’s growing low-altitude economy – which encompasses both commercial drones and emerging flying taxi technology – has been marked as a national strategic priority, with official projections valuing the sector at more than 2 trillion yuan ($290 billion) by 2035. Across many urban and rural regions of China, drones have already entered widespread daily use, supporting applications from commercial food delivery and agricultural crop monitoring to high-rise building exterior cleaning.

Even with this rapid growth, China has emerged as one of the most tightly regulated markets for drone operation globally. With the new rules set to take effect, reports from Beijing indicate that authorized retail outlets for DJI, the country’s largest drone manufacturer, have already begun removing all consumer drones and related components from store shelves ahead of the sales ban.

Official data puts the total number of registered drones across China at more than 3 million, and industry analysts note that the sweeping new restrictions in the capital are expected to reshape the country’s massive domestic drone market, forcing operators and manufacturers to adapt to the new layered regulatory regime.