Global Pulse

分类： technology

  • Apple tries again on AI, turns to Google for help

    Apple tries again on AI, turns to Google for help

    Two years after a high-profile but unfulfilled initial push into artificial intelligence left the tech giant facing public criticism and legal action, Apple has launched a sweeping AI upgrade for its iPhone ecosystem — and it is turning to long-time partner Google to power core capabilities of the new system. Monday’s announcement, made at Apple’s annual Worldwide Developers Conference, also carried historic corporate news: it marked the final keynote appearance from CEO Tim Cook, who will step down from his role in September to hand leadership to veteran hardware executive John Ternus.

    Back in 2022, Cook used the same developer stage to announce Apple Intelligence, a major initiative meant to catapult the company into the global AI race that had already become a top priority for fellow U.S. tech giants. The centerpiece of that announcement was a promised revamp of Apple’s Siri voice assistant, but the overhaul never fully materialized. The unkept promise led to a class-action lawsuit from disgruntled U.S. consumers, which Apple settled earlier this year.

    Unlike many competitors that have poured hundreds of billions of dollars into building custom AI infrastructure and in-house large language models, Apple has adopted a more gradual, cost-conscious approach that has won approval from some industry analysts. At this year’s conference, Apple leaders doubled down on this strategic framing, arguing that rushing to deploy AI for competitive prestige ignores the needs of end users. “AI is incredibly powerful technology with the potential to shape society in profound ways, and with proper care, unlock meaningful benefits for people everywhere,” Apple software chief Craig Federighi said in a pre-recorded launch video. “Still, some appear to be racing forward, seemingly pursuing AI for the sake of AI, without clear regard for the people… that it’s ultimately meant to serve.”

    The centerpiece of Apple’s 2024 AI announcement is a redesigned Siri, rebranded as Siri AI, that will support natural language conversations, cross-app data tracking and task execution across tools including Maps, Mail and other core iPhone applications. Notably, rather than relying on fully in-house developed AI models, Apple has integrated a customized version of Google’s Gemini large language model to power these new features. This partnership builds on an already lucrative existing relationship between the two companies: Google already pays Apple tens of billions of dollars annually to retain its position as the default search engine on iPhone’s Safari browser.

    Most of the AI-powered capabilities Apple unveiled Monday have already been available to consumers through Google, which has rolled out similar generative AI features across its Gmail, Maps and Android operating system products. Prominent Apple tech commentator John Gruber noted that the company’s strategy represents a high-stakes bet on a low-investment AI model. “Apple is making an enormous bet on AI — but their bet is that they don’t need to spend hundreds of billions per year on AI infrastructure…to reap the benefits,” Gruber explained.

    Alongside its AI announcements, Apple used the conference to highlight expanded parental control tools designed to address growing global scrutiny of tech companies over child screen safety and social media addiction. The updated controls include more flexible and robust time management features that specifically target engagement with social media platforms and mobile games.

    Even as Apple lagged behind competitors in rolling out consumer-facing AI features over the past two years, the company’s financial performance has remained strong. Apple stock has defied broader market expectations, rising roughly 15 percent since the start of 2024, and iPhone sales recorded double-digit growth across nearly all of the company’s global markets in the first quarter of the year.

  • Europe opening up to self-driving taxis

    Europe opening up to self-driving taxis

    After years of lagging behind the United States and China, where autonomous robotaxi technology has already scaled rapidly, Europe is finally gearing up to open its doors to large-scale trials of self-driving taxis, with a landmark regulatory shift set to be formalized this week.

    Data from a May International Energy Agency (IEA) report underscores just how far Europe has fallen behind: the total combined fleet of commercial robotaxis in the US and China more than doubled in 2025 alone, hitting 8,000 vehicles operating across more than 25 major cities. Seven years behind the initial global testing timeline, the first wave of multi-city trials is set to launch across the continent over the coming months, with industry leaders from North America, China and Europe already lining up operations.

    To clear the path for faster deployment, the European Union is set to approve a streamlined new “testbed” framework on Monday, senior EU official for autonomous vehicle development Anne-Marie Idrac confirmed to AFP. Currently, strict EU rules require a human safety driver to remain in every test vehicle, a mandate that mirrored early-stage regulations in the US and China. The new framework will eliminate the need for companies to seek separate, country-by-country approval for trials, cutting through bureaucratic red tape that has slowed progress to date.

    The first European trial already launched on April 8 in Croatia, where leading Chinese autonomous vehicle firm Pony.ai has partnered with US ride-hailing giant Uber and Croatian startup Verne, backed by automaker Rimac Group, to operate a fleet of 10 robotaxis through the capital city of Zagreb.

    Multiple more trials are scheduled to launch across the continent before the end of the year. In London, three separate projects are in the works: global robotaxi leader Waymo, a subsidiary of Google-parent Alphabet, will launch its service, followed by UK-based competitor Wayve (in partnership with Uber) and Baidu’s autonomous ride-hailing brand Apollo Go, one of the world’s largest robotaxi operators. In Madrid, Chinese autonomous firm WeRide has just announced a new trial with Uber, while Uber will also roll out test services in Munich powered by Chinese technology firm Momenta’s autonomous driving system. Apollo Go has also paired with Swiss Post for a pilot program in eastern Switzerland, and multinational automaker Stellantis has joined forces with Pony.ai for a trial in Luxembourg. Leading ride-hailing platforms including Uber, Lyft and Bolt have become the most common partners for these cross-industry testing projects.

    Today, the global robotaxi market is already dominated by major players based in the US and China. Wayme says it operates roughly 3,000 fully driverless taxis across a dozen US cities, a fleet size roughly equal to Apollo Go’s, which operates in 27 Chinese cities as well as Dubai. Pony.ai holds a global fleet of 1,700 robotaxis and has set a target of 3,500 vehicles by the end of 2026, while WeRide currently has 1,000 vehicles on roads. Additional key players include Chinese ride-hailing giant Didi and state-owned carmaker SAIC, which operate robotaxi services across major Chinese cities, as well as Tesla and Amazon-owned Zoox, which have deployed test fleets across multiple US cities.

    Industry forecasts project massive growth for the global robotaxi market over the next decade. By 2035, the IEA predicts the global fleet will reach between 700,000 and 3 million vehicles operating across 40 to 80 major cities worldwide. Consulting firm BCG projects a total global fleet of 3 million robotaxis by 2035, with 850,000 based in China and 350,000 in the United States, leaving just 120,000 in Europe. Investment bank Goldman Sachs takes a more optimistic view of market expansion, projecting a global fleet of around 6 million robotaxis that will create a $415 billion global industry by mid-decade.

    For Europe, two long-standing factors have slowed the rollout of robotaxis to date: strict regional safety regulations and Europe’s deeply established culture of high-quality public transit, according to autonomous vehicle specialist Herve de Treglode. Even so, de Treglode predicts commercial robotaxi services could launch as soon as 2027 in ready markets like London and Madrid. Unlike in Europe, where testing has proceeded in slow, fragmented fits and starts, de Treglode noted that the US and China have pursued a far more aggressive rollout strategy: companies deploy to one neighborhood at a time, remove safety drivers once testing is complete, and scale commercial services with massive ongoing investment.

    Even as the EU clears regulatory barriers, Europe’s robotaxi rollout stands at a critical crossroads, with a key disagreement over where services should deploy. Private companies are overwhelmingly eager to launch robotaxi services in dense, highly profitable urban centers, but many policymakers are pushing for services to prioritize underserved suburban and rural areas, where robotaxis could fill gaps in existing public transit networks.

    “It’s high time we came up with a strategy,” Laurence Debrincat of the Paris regional transport authority said last month, advocating for investment to expand service outside of dense city cores. Thomas Matagne, founder of French ridesharing firm Ecov, summed up the core choice facing European regulators: “Should we leave the sector to the market, at the risk of concentrating it in densely populated areas? Or should the government invest to roll out (robotaxis) in the general interest?”

  • China can build humanoids at scale. The hard part is finding enough buyers

    China can build humanoids at scale. The hard part is finding enough buyers

    The global humanoid robot sector is undergoing rapid evolution, with Chinese manufacturers emerging as dominant players in mass production and market deployment, even as industry observers warn of persistent gaps between manufacturing capacity and real-world commercial demand. Against a backdrop of aging domestic populations and rising labor costs across major economies, Chinese robotics startups are positioning humanoid robots as a transformative solution for everything from industrial logistics to consumer household services, drawing strong early interest from both public and private buyers.

    China and the United States currently split leadership in the global race to develop this emerging market, which Morgan Stanley estimates could grow to a $5 trillion valuation. While the U.S. maintains an edge in developing the high-performance artificial intelligence that acts as the advanced “brain” for humanoid systems, China leverages its status as the world’s manufacturing hub to lead in mass production capability, hardware supply chains, and the collection of real-world data required to train robot models.

    Chinese startups are already logging thousands of orders for their humanoid products, spanning government agencies, state-owned enterprises, and private commercial clients. Shanghai-based AI humanoid developer Matrix Robotics, led by former Tesla engineer Allan Zhang, counts roughly 1,000 orders for its flagship MATRIX-3 model, a 5.6-foot-tall robot with precision-controlled manipulator hands priced at approximately $99,000 per unit. Zhang revealed at a recent Macao robotics expo that clients include major coffee chains and hospitality groups, and while the firm has only produced a few hundred units to date, it has the capacity to scale output to 5,000 units annually in 2025 if demand holds.

    Shenzhen-based EngineAI offers a more affordable full-sized humanoid model, with a basic configuration priced at 180,000 yuan ($26,600), targeted at roles including security patrols, museum tour guides, and public performance displays. The company’s brand and marketing head Issac Li says the firm’s next core goal is to expand deployment into more diverse everyday operational scenarios. Leading Chinese manufacturers AGIBOT and Unitree already dominated global shipments in 2025, with each delivering more than 5,000 units that year, compared to just a few hundred or fewer from top U.S. competitors like Figure AI and Tesla. Unitree, one of the sector’s standouts, reported 1.7 billion yuan ($250 million) in 2024 revenue and a net profit of over 278 million yuan ($41 million), marking it as one of the few profitable players in the young industry.

    China’s industry growth has also received substantial backing from national policy. The country’s 2026-2030 five-year plan explicitly names humanoid robotics as a key frontier technology to develop, and by 2025, China was home to more than 140 humanoid robot manufacturers producing over 330 distinct models. Data from Barclays shows Chinese humanoid robots already account for roughly 85% of global supply, and Morgan Stanley projects that annual Chinese sales will more than double in 2025 to around 28,000 units. Industry analysts forecast that widespread adoption will drive down average costs from $46,000 in 2024 to roughly $21,000 by 2050, with Chinese models already selling for 20% less than comparable foreign alternatives on average thanks to localized supply chains, with some entry-level models priced below $6,000.

    Despite this rapid production expansion, industry experts and even government regulators warn that significant hurdles remain before widespread commercialization can be achieved. As early as 2024, China’s Ministry of Industry and Information Technology publicly warned of the risk of industry overcapacity and a market bubble, given the slow pace of viable commercial application development. Most analysts agree that current demand lags far behind existing manufacturing capacity, with most humanoid robots still designed for demonstration rather than reliable functional work in unstructured, unpredictable real-world environments.

    “The use cases of these robots are still so limited,” notes Chibo Tang of venture capital firm Gobi Partners, which invests in early-stage robotics startups. “Without the demand and without that scale from the market, these companies are not able to really go into mass production sustainably.” Samm Sacks, a senior fellow at U.S.-based think tank New America focused on Chinese technology, explains that the core economic barriers remain steep: humanoid robots are still costly to manufacture, prone to operational breakdowns, and only able to function reliably in highly controlled environments. “There’s a long way to go to get to a level of functionality where people will actually feel comfortable having them in their homes providing care for elderly or children,” Sacks added.

    While industry leaders identify industrial logistics as the most near-term viable commercial market for humanoids, many factories across China and the globe already rely on cheaper, more specialized non-humanoid robotic arms for repetitive single tasks, reducing immediate demand for full humanoid systems. The challenge of commercialization is not unique to China: startups in the U.S. and Japan have also struggled to secure consistent bulk buyers for their humanoid models in industrial and service settings.

    Beyond cost and functionality, developers also face the hurdle of accumulating sufficient high-quality diverse real-world data to train robots to handle multiple complex tasks. Eric Guo, founder and CEO of Shenzhen-based AI² Robotics, notes that building a large enough dataset spanning a wide range of public and private scenarios will take years of scaling. “The mass production capability in the robotic area is still at the very early stage,” Guo said.

    Even with these challenges, deployment of humanoid robots in real-world Chinese settings has accelerated sharply over the past year, aided by greater public openness to new technology compared to many other markets. Ye Tian, a former Apple engineer and CEO of Chinese robotics startup RoboScience, notes that Chinese consumers are “used to this rapid change in terms of technology,” creating a more receptive market for early-stage testing.

    Industry insiders remain optimistic about long-term potential, even as they acknowledge near-term headwinds. Lian Jye Su of technology research firm Omdia argues that as the technology matures, humanoids will increasingly take over heavy-lifting and repetitive mundane tasks in warehouses, factories, and ports. Matrix Robotics’ Allan Zhang adds that humanoids can also fill critical gaps in dangerous or labor-short sectors, and predicts a massive untapped consumer market for robots that handle household chores across hundreds of millions of Chinese homes. Early consumer tests already hint at both the potential and remaining flaws: Beijing content creator Yang Ning, who recently tested a robotic cleaning helper that could sort shoes, fold laundry, and replace garbage bags, called the demonstration “amazing” but noted the model was still inefficient, too large, and difficult to maneuver in small living spaces.

    Looking ahead, Omdia forecasts that annual shipments of advanced humanoid robots could surpass 1 million units by the early 2030s, as costs fall and functionality improves, turning a niche experimental sector into a global technology staple.

  • Anthropic calls for pause of global AI development

    Anthropic calls for pause of global AI development

    Leading artificial intelligence developer Anthropic has ignited fresh debate over AI governance this Thursday, calling for a coordinated global halt to work on the most powerful frontier AI systems. The San Francisco-based firm, creator of the popular Claude AI model line, argues that cutting-edge models are already showing early warning signs that they could slip beyond reliable human control.

    In a newly published safety report, the company frames a temporary worldwide slowdown of advanced AI research as a net positive for global society. However, it acknowledges a critical caveat: unilateral action by any single firm would be meaningless, as uncooperative competitors would simply accelerate their own development to gain an upper hand.

    “We believe it would be good for the world to have the option to slow or temporarily pause frontier AI development to enable societal structures and alignment research to keep up with the advance of the technology,” the report reads.

    Anthropic stresses that a functional pause requires buy-in from all major AI developers across leading AI-pioneering nations, most prominently the United States and China. For the agreement to hold, it must also be built around transparent, verifiable rules that all parties can enforce, the company adds. Without such a global coordination framework, both private firms and national governments will be forced to make unenviable safety trade-offs while caught between competitive commercial pressures and growing geopolitical rivalry.

    The proposal has already drawn significant pushback from both industry peers and White House officials. Critics argue that Anthropic’s focus on extreme doomsday scenarios overstates near-term AI risks, accusing the firm of using safety concerns as a pretext to slow down rivals and gain a competitive advantage.

    Notably, the White House has already recognized the exceptional capability of Anthropic’s undeveloped Mythos model, which remains out of public reach due to its advanced cybersecurity functions. The model is currently only deployed to a small, carefully vetted group of organizations.

    The road to implementing Anthropic’s proposal is already steep on both policy and industry fronts in the U.S. Many Washington policymakers and Silicon Valley executives have repeatedly warned that a domestic slowdown in AI innovation would cede a decisive strategic advantage to China in what is widely viewed as the defining global technology race of the 21st century.

    In a surprising turn, former U.S. President Donald Trump noted that he recently discussed potential AI safety cooperation with China during a recent visit to Beijing. This week, Trump also signed an executive order mandating a 30-day preliminary government review of the most powerful U.S.-developed AI models before they can be publicly released.

    Drawing a parallel to historical nuclear arms control agreements, Anthropic warns that regulating advanced AI will prove an even greater challenge. Unlike nuclear missile silos, AI training operations can be easily hidden from international inspectors, creating enormous incentive for parties to cheat on any pause agreement by continuing development in secret.

    Looking ahead, the company says it will convene a broad coalition of stakeholders over the coming months: government regulators, independent AI researchers, public safety advocacy groups, and even competing AI firms, all to work out the practical framework for a verifiable global coordination system.

    Anthropic’s call for action is backed by internal company data that confirms AI is already dramatically accelerating the pace of AI development itself. This auto-acceleration creates a dangerous feedback loop that could eventually lead to the long-debated AI research scenario known as recursive self-improvement, the firm warns.

    Recursive self-improvement describes a scenario where an AI system gains the ability to independently modify and improve its own code and capabilities, becoming increasingly intelligent without meaningful human intervention. While Anthropic emphasizes that this scenario has not yet emerged and is not inevitable, the report notes it could arrive far sooner than most governments and societal institutions are prepared to handle it.

    “The evidence suggests that the human role is narrowing at each step in the AI development process,” the company concludes.

  • Five takeaways from Canada’s new AI strategy

    Five takeaways from Canada’s new AI strategy

    Against a backdrop of rising global public anxiety over artificial intelligence’s risks to privacy, personal safety and employment stability, the Canadian government has launched a long-awaited 10-year national AI strategy that charts a clear path for the country’s adoption and governance of the transformative technology. Prime Minister Mark Carney announced the comprehensive C$2 billion ($1.4 billion) plan on Thursday, framing AI’s proliferation as an irreversible shift that is already reshaping core parts of daily life from work and education to social connection.

    One of the strategy’s central pillars is safeguarding Canadian digital and economic sovereignty, a longstanding policy priority that has gained renewed urgency amid shifting cross-border dynamics with the United States. The plan explicitly identifies reducing reliance on foreign AI infrastructure and service providers as a key goal, highlighting that Canadian companies currently store large volumes of sensitive data in overseas jurisdictions, and the federal government already depends on some critical digital infrastructure owned by foreign entities. Carney warned that bad actors could weaponize AI against Canadian interests, prompting two major infrastructure commitments: the development of a secure, world-class public supercomputer accessible to domestic researchers and businesses, and targeted support for building large-scale domestic AI data centers, with a target of drastically expanding national computing capacity by 2030.

    Addressing a long-standing structural challenge for the Canadian AI ecosystem, the strategy prioritizes retaining homegrown AI talent and attracting global skilled workers to the country. For decades, Canada’s close proximity to the massive U.S. tech market has led to a steady brain drain of top Canadian AI innovators, a reality the strategy openly acknowledges as an “uncomfortable truth.” High-profile examples of this trend include Geoffrey Hinton, the Canadian Nobel Prize-winning researcher known as the “Godfather of AI,” who sold his startup to Google and spent years working for the U.S. tech giant, and Ilya Sutskever, another Canadian AI pioneer who co-founded industry leader OpenAI. To reverse this outflow, the plan will fund new AI research fellowships and expand the number of specialized AI research chairs at Canadian universities. It also offers accelerated immigration pathways and permanent residency for top AI talent from around the world seeking to relocate to Canada. Additionally, the government is committing C$500 million in targeted investment to domestic AI firms, which will allow the federal government to take equity stakes in emerging Canadian AI companies to support their growth.

    The strategy also aims to dramatically scale AI adoption across all Canadian economic sectors over the next decade. Government data shows only 12% of Canadian businesses integrated AI tools into their operations between mid-2024 and mid-2025, and the new plan sets an ambitious target of lifting that adoption rate to 60% by 2034. To reach this goal, the government is offering C$500 million in financing to help small and medium-sized businesses integrate AI into their workflows, alongside C$50 million in targeted support for content creators to adopt AI tools on their own terms.

    A major focus of the adoption push is upgrading Canada’s struggling public healthcare system, which has long grappled with prolonged emergency room wait times and widespread shortages of primary care providers. Carney was joined by healthcare workers from a Toronto hospital for the strategy’s announcement, and the plan earmarks C$200 million to improve health outcomes using AI, with a core goal of reducing the heavy administrative burden that pulls clinicians away from direct patient care. Carney noted that nearly three-quarters of European Union member states already use AI-assisted diagnostics to analyze medical imaging and detect disease, positioning Canada as lagging behind on this front.

    To address widespread public mistrust of AI, the strategy prioritizes expanding national AI literacy. Recent government polling shows deep division among Canadians over AI’s impact: just 34% view the technology as beneficial to society, while 36% see it as harmful, and half of all respondents regard AI as an existential threat to humanity. A global study from KPMG and the University of Melbourne also ranked Canada low among developed nations in AI literacy, training and public trust. In response, the government is launching a national AI literacy initiative that will provide free entry-level AI training to all Canadians, delivered in part through partnerships with the country’s network of public libraries.

    On the regulatory front, the government has pledged to introduce new AI legislation to protect consumer privacy and child safety, as well as updates to modernize Canada’s existing online safety rules to account for AI-specific risks. However, the strategy provides no specific details on the content or timeline of these proposed regulatory changes. Calls for new AI safety rules have intensified in Canada this year, after it was revealed that the suspect in a February mass shooting in Tumbler Ridge, British Columbia, used ChatGPT to discuss gun violence months before the attack. OpenAI acknowledged it was aware of the concerning activity but failed to alert law enforcement, prompting a public apology from CEO Sam Altman and a meeting between OpenAI executives and Canadian officials, who threatened new regulation if the company did not update its safety protocols. Carney emphasized that Canada must be transparent about AI’s risks, including the spread of deepfakes, unregulated unsafe chatbots, and AI-generated disinformation.

    The lack of specific details on safety regulation has drawn criticism from Canada’s Conservative opposition. Conservative MP Melissa Lantsman told reporters that the safety and security safeguards promised by the government are nowhere to be found in the strategy document, with no concrete details on how new rules will work or when they will take effect. While the strategy projects that scaling AI across sectors will create 250,000 new jobs over the next decade, it notably does not include any estimate of how many existing roles could be displaced by rapid AI adoption, leaving a key question about the technology’s labor market impact unaddressed.

  • UK lawmaker says she is suing Elon Musk’s company over fake Grok bikini images

    UK lawmaker says she is suing Elon Musk’s company over fake Grok bikini images

    LONDON — A landmark legal challenge targeting artificial intelligence accountability has emerged in the United Kingdom, as a sitting Labour Party legislator has launched a privacy invasion lawsuit against Elon Musk’s AI development firm xAI over deepfake explicit images generated without her consent by the company’s Grok chatbot.

    Jess Asato, who serves in the UK parliament for the governing party, revealed Thursday that the unauthorized deepfake content was produced in January, shortly after she publicly spoke out against the growing proliferation of non-consensual deepfake pornography across digital platforms. According to Asato’s account, an anonymous party leveraged Grok’s image generation capabilities to create fake photos of her wearing a bikini that were never shot or authorized by her.

    The formal legal claim was submitted to London’s High Court this week, with Asato arguing that xAI violated the UK Data Protection Act through the misuse of her private personal information. Beyond seeking monetary damages for the harm she has endured, the lawmaker has a larger strategic goal: to establish a binding legal precedent that holds AI developers legally responsible for dangerous design flaws in their systems that enable harmful misuse.

    In a statement explaining her decision to pursue legal action, Asato drew a parallel between the online violation and a physical offense. “Nobody would be able to walk up to me in the street and strip me and put me in a bikini, and I don’t see why anybody should be able to do that to me online, because the feeling, while it is not quite the same, is very similar,” she said. “It is like somebody has digitally stripped me without my consent.”

    Asato also said she encourages other people who have suffered similar harm from AI-generated non-consensual deepfakes to join her legal action, framing the case as a broader fight for digital privacy safety.

    This lawsuit comes amid a growing global backlash against the spread of non-consensual deepfake pornography, which has triggered widespread calls for tighter regulation of AI tools. Back in January, after the incident involving Asato drew public attention and international outcry, xAI announced it would update Grok’s policies to ban users from editing images of real people to remove clothing.

    The UK passed a national law last year that explicitly criminalizes the creation or solicitation of non-consensual deepfake images of adults, but Asato argues that existing accountability frameworks are incomplete. Even after companies patch dangerous flaws in their AI systems, she notes, irreversible harm has already been done to victims of misuse.

    “Once the damage is done, the damage is done,” Asato said. “If you think about any other products, like a car, for example, that might have been manufactured with a fault, it doesn’t matter if, you know, the cars get recalled and the faults are fixed and no more harm is done.” Companies must still be held responsible for the harm their flawed products caused before the fix, she argues.

    As of Thursday, xAI had not issued any immediate public response to requests for comment on the new lawsuit.

  • European Union launches tech sovereignty initiative to boost chips, cloud and AI at home

    European Union launches tech sovereignty initiative to boost chips, cloud and AI at home

    BRUSSELS – Growing anxiety over excessive European reliance on non-domestic technology providers has spurred European Union leaders to launch an ambitious new initiative designed to build up homegrown alternatives to foreign Big Tech and critical hardware. On Wednesday, the 27-nation bloc announced its comprehensive “tech sovereignty” package, a set of policy measures crafted to nurture local European competitors for the U.S.-dominated AI and cloud computing sectors and reduce heavy dependence on Asian microchip manufacturing.

    The push for greater technological autonomy has gained urgent momentum in recent years, as EU policymakers have raised alarms that foreign control of critical digital and hardware infrastructure could be weaponized against European interests. These concerns were solidified by a high-profile incident several years ago, when the Trump administration imposed sanctions on the International Criminal Court’s top prosecutor, prompting U.S. tech giant Microsoft to terminate the prosecutor’s corporate email account. The episode fueled widespread fears across the bloc that foreign technology services could include hidden “kill switches” that would allow external powers to disrupt critical European operations at will.

    “Europe wants to be in the position to make its own choices, avoiding risky dependencies on single dominant suppliers, one company or one third country,” Henna Virkkunen, European Commission Executive Vice-President responsible for overseeing the bloc’s tech sovereignty agenda, told reporters in Brussels. “Because we live in a world where geopolitics and technology go hand in hand. Those who champion technological innovation will shape the future, and we must ensure that Europe plays a leading role in this.”

    A core pillar of the new package is an expansion of the EU’s landmark 2023 Chips Act, which was originally introduced to boost local semiconductor output. The updated rules will further streamline burdensome regulatory red tape for new semiconductor fabrication plants, while working to build a fully integrated, self-sustaining European chipmaking ecosystem. These measures were prompted in part by a 2023 power struggle at Nexperia, a Chinese-owned chipmaker headquartered in the Netherlands, which laid bare how vulnerable Europe’s fragmented semiconductor supply chain is to global geopolitical shifts, given most of the world’s advanced chip manufacturing is concentrated in East Asia.

    The second major component of the initiative focuses on shoring up Europe’s domestic cloud infrastructure and artificial intelligence development capacity. The bloc has laid out a formal target to triple its total regional data center capacity over the next five to seven years, a move intended to keep pace with the explosive global AI boom that has driven a sharp, sustained surge in demand for high-capacity cloud computing services.

    The policy package was drafted and released by the European Commission, the EU’s executive branch. It now moves to the bloc’s two other governing institutions — the European Parliament and the Council of the European Union — for debate and final approval before any measures can go into effect.

  • UK orders Google to allow publishers to opt out of AI scraping for search summaries

    UK orders Google to allow publishers to opt out of AI scraping for search summaries

    In a groundbreaking, first-of-its-kind regulatory action announced Wednesday, the United Kingdom’s Competition and Markets Authority (CMA) has ordered Google to grant all online content publishers serving British users the explicit right to opt out of having their work scraped to train the tech giant’s artificial intelligence systems and power its AI-driven search features. The ruling marks one of the most significant regulatory interventions to date to balance the rapid growth of generative AI against the intellectual property rights of content creators, and it uses the CMA’s new digital market oversight authorities to curb what regulators frame as the outsized market power Google holds over the UK’s online search ecosystem.

    Under the terms of the order, Google is required to build and roll out robust, effective tools that let publishers block their content from being used to develop two of the company’s high-profile AI offerings: AI Overviews, the AI-generated summary panels that appear at the top of search results, and the broader AI Mode search experience. Beyond the opt-out right, Google must also implement clear, prominent attribution for any publisher content included in AI-generated search results, with direct working links directing users to the original source material. The order also extends the opt-out right to content used for fine-tuning Google’s large language AI models, giving publishers full control over whether their work contributes to the company’s AI development.

    This ruling was widely anticipated by industry observers, after the CMA released draft proposals for the new rules earlier this year. The regulatory move followed an investigation that confirmed a tangible negative impact on news publishers: after Google launched its AI Overviews feature, publishers saw measurable drops in referral traffic from search, as fewer users click through to original content when an AI summary is provided directly in search results. The new requirements also apply to the sweeping AI updates Google unveiled for its search platform in May 2024, which integrate artificial intelligence more deeply into every layer of the user search experience.

    For Google, the company has signaled it is cooperating with the CMA’s order. In an official blog post, Mrinalini Loew, Google’s general manager for search ecosystem, noted that the company is already working alongside global regulators to give website owners appropriate control over their content as AI reshapes user preferences. “Today, we’re beginning to test a new control that lets website owners manage how their links and content appear in generative AI Search features,” Loew wrote.

    CMA Chief Executive Sarah Cardell framed the ruling as a win for both content creators and UK consumers. The new measures will deliver “fair treatment, greater transparency and meaningful choice for businesses and consumers,” Cardell said, adding that the rules will help tens of millions of British users “better understand and trust the information presented to them.” Regulators also note the order will strengthen publishers’ negotiating position when they enter into content licensing deals with Google, leveling a playing field that has long been tilted toward the U.S.-based tech giant. For the purposes of the ruling, any entity that publishes online content accessible to users in the UK qualifies as a covered publisher, meaning the opt-out right applies to everyone from individual bloggers to large national news organizations.

    The landmark decision sets a global precedent for AI regulation, as other countries around the world grapple with how to address the widespread scraping of copyrighted content to train commercial AI systems. As the first major regulator to mandate a broad opt-out right for publishers in AI search, the CMA’s action could serve as a template for future policy and regulatory action in other major markets.

  • ’20 minutes of terror’: AI boosts US voice impersonation scams

    ’20 minutes of terror’: AI boosts US voice impersonation scams

    Across the United States, a new wave of devastating cyber fraud is taking hold, powered by rapidly advancing artificial intelligence that can replicate a human’s voice with unsettling, near-perfect accuracy. For victims like Buffalo, New York, resident Liz Benz, the experience remains traumatic even after the scam is exposed.

    Benz, a 46-year-old insurance broker and mother of six, still remembers the gut punch of answering an unknown number one afternoon. On the line was a voice that was undeniably her 16-year-old son Fred, crying and shaken, begging for help. The caller told Benz that Fred’s friend had been shot dead, and Fred, who was attending a local high school football game, was being held hostage. To secure his release, Benz was ordered to withdraw a large sum of cash and drop it off at a nearby Walmart parking lot.

    For 20 gut-wrenching minutes, Benz believed her son’s life was in danger. It was only when Fred sent a smiling selfie from the stands of the game that she realized she had fallen victim to an elaborate AI-powered scam. “Nothing could have prepared me to hear my son’s voice, and nothing could have convinced me that this was a scam until I saw my son with my own eyes,” Benz told AFP in an interview, her voice still shaking with the memory of the ordeal. “It was a good 20 minutes of terror.”

    Benz is far from alone. As accessible AI voice cloning tools have become widely available online, U.S. law enforcement and consumer protection advocates are sounding the alarm over a sharp rise in family impersonation scams that exploit emotional panic to steal thousands of dollars from victims. According to Federal Bureau of Investigation data released in April, U.S. consumers lost more than $893 million last year to AI-enabled fraud schemes, including voice cloning scams.

    What makes this threat particularly widespread is how easy it has become for even inexperienced cybercriminals to carry out these attacks. A simple internet search pulls up dozens of free voice cloning applications that can generate a hyper-realistic replica of a person’s voice using just a few seconds of original audio, which scammers can easily harvest from public social media posts, voicemail recordings, or online videos.

    “It used to be somewhat hard to make these things. Now anyone can do it in seconds,” explained Brian Long, chief executive of Adaptive Security, a firm that specializes in AI fraud awareness training. “One guy in a room with a keyboard can make an infinite number of attackers. AI tools can build entire scam scripts off of short snippets of audio captured from public online content.”

    The standard scam script follows a predictable, emotionally manipulative pattern: scammers place an urgent call claiming the target’s loved one has been arrested, injured in a car crash, or involved in a violent crime, and demands immediate payment to resolve the situation. To add credibility, scammers often layer in additional AI-generated voices impersonating police officers, attorneys, court clerks, or bank employees, creating a chaotic narrative that pushes victims to act before they can verify the story.

    Cybersecurity experts note that scammers do not even need a perfectly replicated voice to succeed. A convincing distressed voice that triggers an immediate emotional reaction is often enough to bypass a victim’s critical thinking. “A distressed voice saying ‘mom, help me’ or ‘dad, I’ve been in an accident’ may only need to sound believable for a few seconds,” said Amit Gupta, vice president of product management at cybersecurity firm Pindrop. “The objective is not perfect voice replication. The objective is creating enough emotional uncertainty and urgency that the victim acts before verifying.”

    Since Benz went public with her story, she has received hundreds of messages from other scam victims, most of whom choose to remain anonymous out of feelings of shame and embarrassment. Vulnerable populations, particularly elderly Americans, are disproportionately targeted in what have become known as “grandparent scams”, where scammers impersonate a grandchild in crisis.

    FBI data shows that Americans over the age of 60 reported more than $7.7 billion in total fraud losses last year, a significant jump from 2023. For 73-year-old Philadelphia attorney Gary Schildhorn, who fell victim to a similar scam in 2020, the experience was equally shocking. Schildhorn received a call from an AI-generated voice impersonating his son Brett, who claimed he needed immediate bail money after a drunk driving arrest. Schildhorn rushed to his bank to withdraw the funds, only to get a call from his real son mid-transaction, alerting him to the scam.

    “I will go to my grave swearing that it was your voice, it was your cadence, it was words you would use. There was no accent. It was you on the phone,” Schildhorn told AFP, recalling his conversation with his son after the incident. Like Benz, Schildhorn now partners with Adaptive Security to raise public awareness of the growing AI scam threat, and testified before the U.S. Senate in 2023 about his experience to push for stronger consumer protections.

  • Microsoft testing wearable AI gadget aimed at office workers

    Microsoft testing wearable AI gadget aimed at office workers

    In a revealing showcase at its annual developer conference, Microsoft has pulled back the curtain on two experimental AI-integrated hardware concepts designed to reimagine how professional workers interact with artificial intelligence tools on a daily basis. Unveiled by Microsoft executive Steven Bathiche, the two prototypes represent the tech giant’s latest bet on new device form factors, a direction spearheaded by CEO Satya Nadella under the broader Project Solara initiative.

    The first concept is a compact, desktop-friendly cube device built with both touch and voice-activated controls, tailored to keep AI assistance accessible at a user’s workstation. The second, far more unconventional design, is a wearable AI-enabled access badge that can be strung on a neck lanyard or clipped to a belt loop, delivering on-the-go access to AI-powered work support. Currently, only a few hundred Microsoft employees are testing the prototype devices internally; the company has not confirmed a commercial release date, noting that these early trials will shape how future AI hardware iterations are developed.

    This project marks Microsoft’s latest return to the consumer and enterprise wearable hardware space, following the high-profile HoloLens mixed reality headset line. First launched nearly a decade ago, HoloLens gained major traction when the company secured a multi-billion dollar contract to supply the headsets to the U.S. Army. However, persistent technical issues encountered during military testing led Microsoft to discontinue production of the device line in 2024. Microsoft is far from alone in revisiting the wearable category: Google, which saw its first Google Glass smart glasses launch flop over a decade ago, recently announced it would make a second attempt at AI-powered smart glasses.

    In demonstration footage released by the company, office-based workers were shown interacting with both devices to connect with AI agents – semi-autonomous AI bots that handle a range of work tasks, from drafting to software development assistance – without needing to open a laptop or desktop interface. The lightweight wearable badge, for example, is designed specifically for quick agent interactions when a user is away from their desk. Nadella himself even appeared in footage wearing the badge on a neck lanyard, identical in form factor to the standard employee ID badges worn in most corporate offices.

    The wearable badge also includes an integrated small camera, a feature that has already raised familiar privacy questions. During his conference presentation, Bathiche demonstrated the camera by activating the device via fingerprint sensor, pointing it at the conference audience to capture photos, and directing the AI agent to send the images to his personal device for review. In a public blog post, Bathiche explained that the built-in camera is intended to help AI agents interpret the user’s surrounding environment and take contextually relevant actions. However, camera-equipped AI wearables have faced intense public and regulatory scrutiny in recent years; Meta’s AI-powered smart glasses, for example, have drawn widespread criticism over a lack of transparency around when and how user footage is recorded, stored, and used. This latest prototype from Microsoft is likely to face similar scrutiny as it moves through internal testing.

    As AI agent technology becomes increasingly central to the workflows of software developers and other knowledge workers, major tech leaders have framed AI automation as a core driver of recent corporate layoffs that have displaced tens of thousands of workers. These new concepts from Microsoft signal the company’s broader push to embed AI agent functionality into every layer of daily work, beyond traditional computing devices.