After nearly seven years at the helm of Meta’s globally dominant messaging platform WhatsApp, Will Cathcart has announced his departure from the top role, in one of the most high-profile leadership changes in Big Tech in recent months. During Cathcart’s tenure, WhatsApp expanded exponentially, growing its global user base of private messaging to over 3 billion people, while rolling out new encrypted communication features that cemented its position as the world’s most widely used mobile messaging service. In social media posts shared Monday, Cathcart noted that WhatsApp is currently in the strongest strategic position in its history, making the timing right for him to step back from day-to-day leadership of the platform. While Cathcart will leave his role as WhatsApp head, he will remain a member of Meta’s senior leadership team. Taking his place is Kunal Shah, the founder of Bangalore-based Indian fintech unicorn Cred, a fintech firm that has shaken up India’s credit and payments sector since its launch. Meta chief executive officer and Facebook co-founder Mark Zuckerberg voiced strong public support for Shah’s appointment this week. Zuckerberg highlighted that Shah built Cred into one of India’s most consequential technology companies, and said his builder-focused mindset and global outlook make him uniquely suited to lead the world’s largest messaging platform. “I look forward to working with Kunal to continue to make WhatsApp the best service for billions of people and millions of businesses,” Zuckerberg said. Shah founded Cred in 2018, building a members-only fintech platform that rewards users for making on-time credit card payments, with a core focus on high-income users in India’s fast-growing digital economy. Before launching Cred, Shah built a career as an investor and advisor to early-stage startups across India and Southeast Asia, according to his public LinkedIn profile. In his own announcement shared on social media Monday, Shah confirmed he will retain his stake as a shareholder in Cred while taking on the WhatsApp leadership role. He also revealed that Cred has secured $900 million (approximately £679 million) in new investment from Meta, a deal that Bloomberg reports will give Meta a 20% minority stake in the fintech firm. To address growing privacy concerns that have dogged Meta in key emerging markets, Shah emphasized that while Meta will hold a minority stake in Cred, the social media giant will not gain access to any of Cred’s member user data. The leadership reshuffle at WhatsApp comes as Meta works to deepen its foothold in India, WhatsApp’s largest single market by user count. Data from the World Population Review shows that WhatsApp counts roughly 853 million active users in India alone, accounting for more than a quarter of the platform’s total global user base. As part of Meta’s broader “family of apps” that also includes Facebook, Instagram, and Messenger, the company has increasingly looked to WhatsApp to drive new revenue growth, testing new revenue streams including in-app advertisements, paid premium subscription features, and artificial intelligence-powered tools. Despite its massive popularity in the country, WhatsApp and Meta have faced growing regulatory and public scrutiny in India in recent years over the platform’s privacy policies and data sharing arrangements between the app and its parent company. The leadership change and Meta’s new investment in Cred signals the company’s long-term bet on solidifying its position in the world’s largest open digital market, as competition for users and revenue in emerging technology sectors intensifies across South Asia.
分类： technology
-
Why an AI company cleaned my New York City apartment for free
Across bustling New York City, a unusual new advertising campaign is turning heads: billboards and street displays feature cleaners outfitted with head-mounted cameras, offering city residents a free, no-obligation home clean. What many passersby may not immediately realize is that this offer is the frontline of a bold new AI initiative that is reshaping how robotics companies collect real-world training data, while stoking fierce debate over personal privacy in the age of artificial intelligence.
The project, called Shift, is the brainchild of AI startup Micro AGI. Its premise is straightforward: participating homeowners get complimentary deep cleans and even private chef visits to their residences. In exchange, every member of the service team wears a camera mounted to their cap, wired to a mobile device that records every action taken inside the home — from wiping down kitchen counters to organizing living room shelves to preparing meals. Every inch of the participating property is captured, with all footage anonymized before being packaged as training data for the next generation of autonomous humanoid robots.
Tech industry leaders have long pinned their hopes on versatile autonomous robots that can handle everything from routine household chores like washing dishes to full-time in-home care for elderly and disabled people. The biggest barrier to bringing these robots to market, however, is the lack of high-quality, real-world data to teach robotic hands how to navigate the infinite variation of human homes. Unlike large language models such as ChatGPT, which can train on trillions of words of publicly available text online, physical robots need to learn how different kitchens are laid out, how different cleaning tools work, and how to adapt to changing lighting, object placements and home layouts — all things that cannot be simulated in a lab.
A BBC reporter who participated in the Shift program at their Upper East Side apartment was greeted by two young graduates in their mid-20s who had previously worked in the startup ecosystem and joined Shift for steady work. With demand for the free cleaning service far outstripping the company’s capacity, the team is permanently stationed in New York, cleaning up to five homes per day, five days a week. During the clean, the pair focused heavily on the movements of their hands, as every dexterous action is exactly what the AI model needs to learn from.
Shift founder Bercan Kilic told the BBC that the company’s entire mission is aimed at advancing AI capability for the public good. “In the real world, every object is different, the lighting is different and nothing is the same as it was a couple of hours earlier,” he explained. “Models need to learn how their hands, cameras and environments work together.” To reach that goal, the company needs to collect massive volumes of data — a hurdle the free service model is designed to solve. Kilic added that Shift’s long-term vision extends far beyond residential cleaning: the company already collects data from mechanics performing car repairs in Turkey, and eventually hopes to record nearly every skilled task humans perform, offering free or discounted services in exchange for the data. Its business model centers on selling the anonymized datasets to robotics firms and other AI developers working on commercial humanoid robots.
Not everyone, however, is welcoming Shift’s approach. Leading data privacy and human rights advocates have sounded the alarm over the model of trading free services for access to the most private space a person has: their own home. Rory Mir, director of open access and tech community engagement at digital rights group Electronic Frontier Foundation, said the initiative is part of a growing worrying trend of “data bribes” that trade small short-term benefits for long-term privacy risk. “While it might come with money or a service upfront, the data you share has a way of coming back to bite you,” Mir warned. “Even if you trust the business collecting it, there is always a risk of them sharing that information with other businesses or governments. We have just lived through decades of our data being used to manipulate us with advertising and predatory practices like surveillance pricing.”
Calli Schroeder, director of the AI and human rights program at the Electronic Privacy Information Center, went even further, calling Shift’s model “a diabolically creative way to sell privacy invasion.” Schroeder pointed out two overlapping risks: first, the sensitive personal information captured by in-home recordings is far more detailed than most people realize, and the profits from selling that data dwarf the small value of a free clean. Second, the end product of this data collection — fully autonomous cleaning and household robots — could ultimately displace millions of working people who rely on household service jobs for income.
Kilic pushes back against these criticisms, arguing that Shift is far more transparent than the countless other companies that collect user data every single day without users’ knowledge or consent. “Clearly your data is being used every single day, but you don’t know what for and you are not being paid,” he said. “But a free service means at least you are being paid, and it is as honest and as transactional as that. If you don’t want to do it, you don’t have to. We don’t expect everyone to like it and that is fine.”
Despite the privacy concerns, the model has found willing participants. Many New York residents have jumped at the chance to get a free clean while contributing to cutting-edge AI development, and even Shift’s own cleaning staff are enthusiastic about the project. The team, made up largely of Gen Z workers, are paid above the standard cleaning rate in New York, and many see themselves as early pioneers in the coming AI revolution. One cleaner even sent a recording kit home to his own mother, who now records herself doing daily chores around her house to contribute to the dataset. For proponents, Shift is a transparent, mutually beneficial model that accelerates the development of transformative AI technology — for critics, it is a dangerous new frontier of privacy exploitation that trades long-term personal security for short-term convenience.
-
Nepal foreign minister looks to AI partnership opportunity with China
In a recent statement highlighting growing cross-regional technological cooperation, Nepal’s top diplomat has thrown his support behind deepening artificial intelligence collaboration with China, framing the partnership as a key catalyst for Nepal’s own technology-focused national development.
Speaking ahead of planned diplomatic engagements focused on bilateral technological exchange, Nepalese Foreign Minister Shisir Khanal praised China’s dramatic, world-leading progress in the global artificial intelligence sector over the past decade. He emphasized that Nepal sees significant untapped potential for mutual benefit in working alongside China to grow its domestic tech ecosystem, rather than pursuing an isolated development path.
Khanal noted that AI-driven innovation could unlock major gains across Nepal’s core economic sectors—from agriculture and tourism to small and medium enterprise development and public service delivery. By partnering with China, a global leader in AI research and commercial AI deployment, Nepal can accelerate its transition to a more digitally advanced, technology-driven economy aligned with its own national development priorities, he added.
The announcement comes as demand for AI capacity-building and partnership grows across low- and middle-income South Asian nations, many of which are looking to integrate emerging digital technologies to boost inclusive growth. China has already expanded a range of technological cooperation initiatives across South Asia through frameworks like the Belt and Road Initiative, opening new opportunities for knowledge sharing, infrastructure development, and joint research in cutting-edge fields including artificial intelligence.
-
Federal regulators back Trump’s plan to speed power to energy-hungry AI data centers
WASHINGTON – In a landmark move tailored to the explosive growth of artificial intelligence, federal energy regulators have greenlit new rules to speed up connections between large power users — most critically AI-focused data centers — and the nation’s aging electric transmission system. The vote comes as U.S. policymakers race to meet surging energy demand for AI infrastructure and shore up American competitiveness against China in the fast-expanding tech sector.
The policy shift was initiated eight months ago, when U.S. Energy Secretary Chris Wright called on the Federal Energy Regulatory Commission (FERC) to take bolder action to clear interconnection bottlenecks that have delayed data center development. In a unanimous vote by the five-member commission, FERC passed an order requiring that large power loads such as AI data centers gain timely, orderly access to high-voltage transmission lines. This marks the second major regulatory adjustment FERC has made in less than a year to accommodate AI infrastructure, following a December order that allowed data centers to connect directly to independent power plants.
FERC Chair Laura Swett, a Trump administration appointee, framed the decision as a historic step to modernize the U.S. electricity market while shielding ordinary consumers from unexpected costs. Under the new order, data center operators are required to cover 100% of the costs for transmission upgrades needed to support their connections, addressing widespread concerns that general ratepayers would be forced to foot the bill for expanding the grid to serve AI facilities. “Many Americans are increasingly concerned that adding large power loads like data centers will push up their electricity bills, and that concern is shared by this commission,” Swett said. “We take very seriously our congressional mandate to ensure rates remain reasonable for American households.”
The policy has drawn a mixed reception from stakeholders across the energy and political landscape. Tech giants and data center developers have welcomed the reform, which cuts through years-long interconnection backlogs that have left major AI companies scrambling to secure enough power to expand their operations. Major industry players including xAI, Google, Microsoft, Meta, Oracle, OpenAI and Amazon have already backed the administration’s efforts by signing the Ratepayer Protection Pledge, a voluntary commitment to build or purchase new power generation to match their data center demand, cover all grid upgrade costs, provide emergency backup power to avoid blackouts, and prioritize local hiring for construction projects.
But the rule has sparked significant pushback from utilities, state governments, and regional grid operators, who argue that the new federal framework erodes their existing authority to manage local transmission planning and interconnection processes. Clean energy advocates have also raised alarms, warning that the policy could undermine state-level requirements to power new data centers with renewable energy sources, slowing progress on national decarbonization goals.
Beyond regulatory and policy disputes, the FERC vote comes amid a growing grassroots backlash against data center development across the country. Local communities are increasingly pushing back against proposed facilities, citing fears that concentrated data center demand will drive up regional electricity prices, strain local water supplies, and cause environmental harm. Many residents have also protested the loss of open space and farmland to massive data center complexes, which have grown dramatically in size to meet AI’s extreme power requirements.
Current industry data underscores the scale of AI’s growing energy footprint. An estimated 4,000 data centers are already operating across the U.S., with another 3,000 planned or under construction — some of which consume as much electricity as a small entire city. The Electric Power Research Institute estimates that data centers already account for roughly 5% of total U.S. electricity demand, a figure that could triple by 2035. In Virginia, a major hub for data center development, facilities already make up more than 25% of the state’s total electricity demand, a share projected to climb above 40% by 2030.
Even with the new FERC rules, however, major structural challenges remain to matching AI’s power demand with available supply. The order does not address the growing gap between data center construction and the buildout of new power generation plants. In many regions, data center development is outpacing the addition of new generation capacity, leading to tightening energy supplies, rising retail electricity prices, and increased warnings of potential future blackouts.
Industry analysis also shows that even with clearer interconnection rules, overall data center capacity buildout is lagging far behind projected demand. A recent J.P. Morgan report, which analyzed satellite imagery of planned projects, found that more than 60% of data center capacity scheduled for completion by 2027 has not yet broken ground, and an additional 7% of projects have already been delayed. The report cited persistent permitting bottlenecks, supply chain delays for critical equipment including gas turbines and transformers, and widespread shortages of skilled construction labor as the primary causes of backlogs.
For the Trump administration, accelerating AI data center development is a core priority to maintain U.S. economic and military leadership in AI, a sector the White House views as critical to outcompeting China. President Trump has made AI competitiveness a central policy focus, recently signing an executive order to establish a 30-day national security vetting framework for the most advanced AI systems before they can be released to the public, while framing the sector as a magnet for global foreign investment into the U.S.
-
AI executives gather at G7 as Europeans seek checks on American dominance
On Wednesday, some of the world’s most influential artificial intelligence executives convened in France, capping off the Group of Seven major industrialized nations summit with a urgent conversation centered on balancing AI innovation with growing global demands for technological independence from U.S. industry dominance.
While this year’s G7 summit was dominated by discussions of ongoing armed conflicts in Iran and Ukraine, the final day of the gathering carved out dedicated space for one of the most pressing technological issues of our time: the future of global AI governance and development. In a rare high-profile gathering of cross-border AI leadership, the chief executives of three of the world’s most powerful AI companies – OpenAI’s Sam Altman, Google DeepMind’s Demis Hassabis, and Anthropic’s Dario Amodei – gathered for a working lunch focused on the goal of “Ensuring a safe, rapid and effective deployment of artificial intelligence.”
The meeting was not limited to the sector’s largest players: it also included the heads of emerging AI labs from across the globe, including Canada’s Cohere AI, French developer Mistral AI, Germany’s Black Forest Labs, Italy’s Domyn, Japan’s Sakana AI, and U.K.-based generative AI firm Synthesia.
This summit comes amid a sharp rise in calls for tech sovereignty across Europe and other non-U.S. regions, driven by mounting concerns about the outsized control U.S. companies hold over the global AI ecosystem. Just weeks before the G7 gathering, the European Commission rolled out a sweeping tech sovereignty strategy aimed at accelerating the growth of homegrown European AI development. Even the Pope added his voice to the debate last month, calling for strict, globally coordinated regulation of artificial intelligence to prevent unchecked domination by a small number of major powers.
Tensions around this issue flared just one week before the summit, when Anthropic was forced to take its two most advanced AI models, Claude 5 2 (wait correction original it’s Fable 5 and Mythos 5, right) – Fable 5 and Mythos 5 – offline globally to comply with an executive order from the Trump administration that cited unspecified U.S. national security priorities. The order barred all non-U.S. persons, regardless of their location, from accessing the models, forcing the company to cut off access for every international customer overnight.
That sudden blackout served as a stark wake-up call for governments and industry leaders around the world, highlighting the extreme strategic vulnerability that comes from relying on foreign-controlled AI infrastructure. Zach Meyers, research director at Brussels-based think tank CERRE, noted that the incident laid bare just how exposed non-U.S. nations are to unilateral policy shifts from Washington. “There is a general anxiety about the state of Europe, the fact that we’re relying on other countries for quite important strategic infrastructure and a desire to do something about it, whatever that is,” Meyers explained.
Canadian Prime Minister Mark Carney echoed that concern on his way to the G7 summit, speaking to reporters during a stop in Dublin. The Anthropic incident, he said, makes clear the urgent need for the global community to “build out and diversify” AI development capacity. True national sovereignty, Carney emphasized, depends on “unhindered access to AI” that cannot be cut off by the policy decisions of another country. Earlier this month, Canada unveiled its own plan to help middle powers and like-minded nations develop alternative AI ecosystems independent of the largest U.S. players. The move came just days after the Trump administration released an executive order outlining a new framework for oversight of cutting-edge AI systems.
For host nation France, the conversation around digital AI sovereignty is far from new: French President Emmanuel Macron has made the issue a core policy priority for years, even mandating that French civil servants replace U.S.-owned video conferencing tools Zoom and Microsoft Teams with a domestic French alternative.
Aidan Gomez, CEO of Canadian AI firm Cohere – which acquired leading German AI startup Aleph Alpha earlier this year – outlined his company’s goals for the summit, saying the firm aims to expand sovereign AI ecosystem partnerships beyond its existing bases in Canada and Germany to include all G7 nations and private sector stakeholders. The end goal, Gomez explained, is to establish a global standard that guarantees national and local ownership of AI models, training data, and computational infrastructure.
In addition to the seven core G7 members – France, the United States, Canada, Germany, Italy, Japan, and the United Kingdom – the summit invited guest nations including Brazil, India, Kenya, and South Korea to participate in select discussions, broadening the global perspective on AI development and sovereignty.
-
AP Exclusive: Nvidia’s Jensen Huang says society needs ‘new social norms’ in the age of AI
SHERMAN, Texas — As the leader of the firm that ignited the global artificial intelligence boom, Nvidia Chief Executive Jensen Huang laid out a comprehensive vision for AI’s integration into modern life Tuesday during an exclusive interview with the Associated Press, arguing that widespread embrace of the transformative technology will deliver broad societal benefits while calling for deliberate adaptation to new norms and targeted regulation.
Huang, whose company’s explosive growth fueled by AI demand has pushed its market capitalization past $5 trillion to make it the world’s most valuable publicly traded company, has long voiced unbridled optimism about AI’s ability to reshape economies and accelerate scientific discovery. But as public anxiety grows over the technology’s potential harms — from mass layoffs to existential safety risks — the industry’s most prominent executive has stepped forward to address critics, pushing back against fears while acknowledging the need for proactive change.
“We need to create new social norms,” Huang said during the interview. “I would advocate that everybody use AI. Just go engage it.”
His remarks come at a moment when AI has become a contentious political flashpoint across the United States. Communities have pushed back against plans to build new AI-focused data centers over environmental and infrastructure concerns, while workers across sectors worry that rapid adoption will leave millions jobless without adequate social safety nets. These growing concerns have eroded public support for the technology even as the U.S. faces intensifying AI competition with China — a race Huang says the U.S. can only win by maintaining an open, globally engaged approach to AI development.
Huang pushed back against narratives that AI will leave non-technical workers behind, noting that the technology has already narrowed the digital divide by enabling people without coding or software development skills to complete complex tasks ranging from website design and dense document analysis to cutting-edge scientific research and home renovation planning.
Drawing a historical parallel to the introduction of automobiles, Huang argued society will adapt to AI just as it adjusted to the new technology of a century ago. “Cars were once portrayed as killing children, but the world changed its norms by having sidewalks and crosswalks and stopping kids from playing in the streets,” he explained, framing current anxiety as a natural part of integrating a transformative new technology.
On the topic of regulation, Huang acknowledged that targeted government oversight and baseline safety standards are necessary, stressing that national security must remain a top priority for a technology that has been a key driver of recent U.S. stock market gains and economic growth.
He voiced skepticism over a recent cross-ideology proposal that would have the U.S. government take equity stakes in AI companies to ensure the public broadly shares in the sector’s windfall profits, an idea floated by former and current President Donald Trump, independent Sen. Bernie Sanders, and even OpenAI CEO Sam Altman. Huang noted that American companies already deliver broad benefits to the public through multiple channels: “Their success benefits the stock price, of which many Americans are investors in. It generates taxes, which helps many Americans. It creates a lot of jobs.” He added that growth in the AI sector also lifts profits for connected industries including energy, construction and hardware manufacturing, meaning Americans already hold a natural stake in AI firms’ success across the economy.
Huang addressed the Trump administration’s recent shift toward stricter AI regulation, including new export controls on Anthropic’s latest AI models that forced the company to suspend public access to the tools, and a new executive order requiring voluntary government screening of high-impact AI models before release. He agreed that national security must be the top priority for all emerging technologies, but called for clear, targeted policy: “you have to be very specific about the risk that you’re concerned about, before setting up policies for export controls.”
This is not Huang’s first run-in with AI export controls: during the Biden administration, Nvidia pushed back against restrictions on chip sales to China, rejecting the argument that a ban would protect U.S. AI advantages. Huang warned at the time that broad restrictions would undermine the development of a global U.S.-led AI ecosystem, as China would respond by accelerating development of its own advanced chips.
Huang also identified energy supply as one of the most critical vulnerabilities for U.S. AI development, noting that AI training and inference data centers require massive amounts of electricity that risks straining the national power grid and raising household utility costs. “The United States is woefully behind in energy production,” he said. “We just suffocated energy production for too long.” Without expanded energy output, he warned, the U.S. will struggle to capitalize on its leading position in AI chips, models and infrastructure. Huang complimented Trump’s policy of expanding domestic fossil fuel production, declining to comment on Trump’s rejection of large-scale solar and wind energy development.
The interview took place during a visit to Sherman, Texas, for the expansion of a Coherent factory building new laser systems that transmit data between chips, a technology that could cut AI power consumption by as much as 50%.
Huang’s close public friendship with Trump has drawn criticism from Democrats, and Huang shed new light on how the relationship began. The pair first connected last year, when Huang was in South Florida to accept the Edison Achievement Award for his work on AI, and Trump invited him and his wife Lori to dinner at his Mar-a-Lago private club. “He was incredibly engaging, incredibly charismatic, conversational, asked a lot of questions,” Huang recalled of the meeting. “From the moment that I met him, the only thing that he’s ever talked to me about is creating more jobs, reindustrializing the United States, protecting national security, winning.” Huang added that Trump often calls him unexpectedly to discuss policy around these priorities, and arranged for Huang to be picked up by Air Force One in Alaska during Trump’s recent state visit to China.
Sen. Elizabeth Warren was among the Democratic critics who attacked Huang for declining to testify before a Senate panel while attending the high-profile Mar-a-Lago dinner, which reportedly had a $1 million per person admission price. Huang pushed back on the political criticism, noting that he supports presidential success regardless of party affiliation: “We could differ with politics, but we should want him to succeed. Because when President Trump succeeds, our country succeeds.”
Huang also reaffirmed that AI’s long-term impact will be overwhelmingly positive, arguing that full engagement with the technology rather than fear-driven restriction will position the U.S. to lead the world in delivering shared growth and progress.
-
Rights groups call for a halt of AI tech use in the military
A broad coalition of more than 100 global human rights organizations and technology activists has issued an urgent warning about the accelerating integration of artificial intelligence into military operations, pointing to Israel’s military campaign in Gaza as a catastrophic and dangerous precedent for the future of warfare. The joint statement, signed by prominent groups including Amnesty International, Access Now and the Stop Killer Robots Campaign, warns that expanding AI’s role in end-to-end military kill chains carries severe risks of escalating civilian casualties and eroding mechanisms for holding perpetrators of violence accountable for human rights violations.
The coalition specifically highlights the widespread deployment of AI-powered targeting systems by the Israeli military during its ongoing operations in Gaza, where more than 73,000 Palestinians have been confirmed killed, with thousands more still trapped and presumed dead beneath the rubble of destroyed residential and infrastructure sites. Three AI tools in particular — named Lavender, Gospel, and Where’s Daddy — have been leveraged to generate air strike targeting lists, drawing on massive Israeli mass surveillance datasets collected on Palestinian residents of Gaza. Rights groups argue that this AI-driven targeting process, which selects targets with little to no active human oversight, is a key contributing factor to the unprecedented scale of death and displacement in the territory.
“The adoption of AI targeting systems in this campaign follows the example of the Israeli government’s weaponisation of data analysis and machine learning tools, powered by mass surveillance, in its genocidal attacks on Gaza,” the statement reads. “By diluting human responsibility for life-and-death decisions, Israel’s use of systems such as Lavender, Gospel, and Where’s Daddy may contribute to the obfuscation of international crimes behind a veneer of perceived algorithmic objectivity while also obfuscating accountability.”
The coalition has issued a clear call to action for global technology companies to immediately end all provision of AI and related technical support to the Israeli military and other state armed forces, warning that unregulated deployment risks normalizing and accelerating the proliferation of AI-powered tools of war. In recent years, many of the world’s largest technology firms have deepened their institutional ties to military establishments, particularly the U.S. Department of Defense (DoD). OpenAI has publicly confirmed it provides AI services to the DoD, while Google holds a DoD contract to “develop prototype frontier AI capabilities to address critical national security challenges across warfighting and enterprise domains.” Microsoft, Google, and Amazon have also all secured contracts to provide cloud data storage, processing capacity, and other core enterprise infrastructure to support the DoD’s warfighting programs.
These corporate military partnerships have triggered significant internal backlash from technology workers, who fear they are being forced to participate in projects that enable human rights abuses. Last month, hundreds of employees at Google DeepMind’s UK artificial intelligence division voted overwhelmingly to unionize, driven by growing concerns over the company’s technology being used by the U.S. for military action against Iran and by Israel against Palestinians in Gaza. UK-based DeepMind staff formally requested that Google management recognize the Communication Workers Union (CWU) and Unite the Union as their official workplace representatives, a move organizers say marks the first unionization drive at a major global frontier AI research lab. In the internal vote among CWU members at DeepMind, 98 percent of participants supported the unionization push. Beyond union representation, DeepMind workers are demanding an end to the use of Google AI by Israel and the U.S. military, the restoration of a previously scrapped corporate commitment not to develop AI-powered weapons or surveillance tools, the creation of an independent ethics oversight body to review high-risk projects, and formal legal protection for workers who refuse to participate in projects on moral grounds.
-
Japan’s tech business SoftBank rolls out OpenAI ‘patches’ against cyberattacks
Two major global technology players, Japan’s SoftBank Group Corp. and U.S.-based OpenAI, announced Tuesday the official launch of a new artificial intelligence-powered cybersecurity service designed to counter the rapidly growing threat of sophisticated cyberattacks across Japan, according to joint statements from both firms.
Speaking at the launch event held in Tokyo, SoftBank’s iconic Chief Executive Masayoshi Son framed Japan’s current cybersecurity gaps as an urgent national crisis, noting that modern cyber threats are far more destructive and widespread than previous generations of attacks. He drew a stark comparison between today’s threat landscape and historical risks, describing contemporary cyberattacks as equivalent to a machine gun assault, versus the far less damaging rifle-level attacks of years past.
The new offering, branded as an AI-powered vulnerability patching service, will first target Japan’s 3,000 largest companies that oversee the country’s most critical public and economic infrastructure — including airport operations, national power grids, and nationwide transportation networks, Son confirmed. For Son, the project is more than a commercial venture: “I feel it is our duty,” he stated, repeatedly labeling malicious cyber actors as “the bad guys” that the new service is built to stop.
The service follows a two-step workflow to shore up defenses: first, it runs a comprehensive diagnostic scan to map unaddressed security weaknesses across an organization’s digital systems, then it leverages OpenAI’s advanced technology to analyze vulnerabilities and generate targeted fixes for these security gaps.
OpenAI CEO Sam Altman, who was originally scheduled to appear in person at the Tokyo event, was unable to attend after his daughter was born earlier than expected. Instead of an in-person talk, Altman delivered remarks via a pre-recorded short video, and OpenAI’s Chief Researcher Mark Chen attended the launch on his behalf.
The partnership behind this new service builds on a collaboration the two firms forged last year, when they launched SB OAI Japan, a 50-50 joint venture focused on building and exclusively distributing customized AI solutions for the Japanese domestic market.
While no financial details of the new cybersecurity service rollout were disclosed at Tuesday’s presentation, SoftBank confirmed that all attendees of the Tokyo event are eligible to apply for a complimentary initial vulnerability diagnosis of their organizational systems.
Industry observers note that the widespread adoption of generative AI by both defenders and attackers has reshaped the cybersecurity landscape in recent years: bad actors have used AI to scale up the volume and complexity of their attacks exponentially, forcing security providers to pivot to AI-powered defense tools that can match the speed and sophistication of modern threats. This new SoftBank-OpenAI offering marks one of the most high-profile commercial launches of AI-powered cybersecurity tailored for a national critical infrastructure network to date.
-
Anthropic to meet White House over AI tool suspension
Top leadership from leading artificial intelligence startup Anthropic will convene with senior White House and U.S. Department of Commerce officials in Washington D.C. this Monday, a gathering prompted by emerging national security risks tied to the company’s recently launched cutting-edge AI models, two anonymous sources familiar with the planned meeting confirmed.
The scheduled discussion comes on the heels of a rapid sequence of events that saw Anthropic halt all public access to its latest AI tool release this past Friday. The pullback followed an explicit federal order barring the company from granting any foreign national access to the advanced technology, which carries far greater capabilities than most publicly available AI systems currently on the market.
The models at the center of the current debate are Fable 5 and Mythos 5, two variants of Anthropic’s next-generation Claude Mythos architecture. Fable 5, the version configured with additional safety guardrails, was released for general public use, while Mythos 5 — which operates under alternative access controls — is restricted to a small curated group of approved organizations. The Claude Mythos line first made headlines back in April, when Anthropic rolled out limited preview and testing access to a small cohort of entities, including multiple U.S. government agencies.
According to insiders, Anthropic CEO Dario Amodei will personally attend the Monday meeting alongside Commerce Secretary Howard Lutnick. Official spokespersons for the White House declined to provide any on-the-record comment when reached for this report, while representatives from both the Commerce Department and Anthropic also did not respond to requests for comment.
Last week, when Anthropic first announced the public rollout of Fable 5, the company openly acknowledged the rollout carried inherent risks. In a public statement, Anthropic noted that Fable 5 outperforms every model the company has ever released for general use, a capability that has drawn heightened scrutiny from federal regulators.
Just days after the public launch, federal authorities flagged that they had identified a potential “jailbreak” vulnerability — a loophole that could allow bad actors to coerce the AI into carrying out functions it was never designed or approved to perform. Anthropic responded Friday that it had only received unsubstantiated verbal reports of the claimed vulnerability, with no concrete evidence provided to date.
The current standoff over the new models marks the second high-profile conflict between Anthropic and the federal government this year. Earlier in 2025, Anthropic filed a lawsuit against the U.S. Department of Defense over disputes regarding permissible use cases for its AI models. Just a few weeks ago, however, tensions appeared to ease after company leaders held a “productive” meeting with senior White House officials, raising hopes that regulatory disagreements could be resolved collaboratively.
Sources familiar with Monday’s planned agenda say the meeting will center on a full formal documentation of the alleged national security and vulnerability concerns raised by federal officials. As of this report, it remains uncertain whether Anthropic will be permitted to restore public access to Fable 5 and limited access to Mythos 5 following the discussion.
-
Dozens of Stanford grads walk out on Google CEO’s speech
In a striking display of public dissent that has drawn widespread attention, dozens of graduating students from Stanford University walked out of the commencement speech delivered by Google’s CEO mid-address. The demonstration, captured on camera by BBC journalists, saw protesters leave the ceremony venue carrying handcrafted signs that highlighted their deep unease over Google’s controversial partnership with the United States government.
The walkout, which unfolded in front of thousands of fellow graduates, guests, and university faculty, spotlights growing tension between big tech firms and younger tech professionals over the ethical implications of corporate work for government agencies. Critics of the partnership have raised a range of concerns, from questions about data privacy and civil liberties to the potential misuse of artificial intelligence and other Google technologies in government programs. Many of the protesting students, who are set to join the tech workforce in the coming months, say they cannot in good conscience stay silent while their future industry leader engages in work they view as ethically compromised.
This demonstration is not an isolated incident: in recent years, Google employees and student tech groups have repeatedly organized walkouts and petitions to protest the company’s business decisions, from workplace inequality to international government contracts. Monday’s walkout at one of the world’s most prestigious universities signals that this grassroots opposition within the tech community continues to gain traction, even as the company maintains its government partnerships and defends its policy decisions.