Per reports from La Tribune and the Financial Times, SoftBank Group Corp. plans to invest as much as €75 billion ($87 billion) to build artificial intelligence data centers in France. The reports present the commitment as aimed at creating large-scale AI compute infrastructure in the country.
NVIDIA unveiled the RTX Spark superchip at GTC/Computex, a Windows-on-Arm platform co-developed with Microsoft, designed for local AI agents and content creation. According to NVIDIA's May 31 press release, RTX Spark combines up to 6,144 CUDA cores, a 20-core Arm-based NVIDIA Grace CPU, and up to 128GB of unified memory, and delivers about 1 petaflop of AI performance. NVIDIA and Microsoft say the platform will ship in thin-and-light laptops and compact desktops from partners including ASUS, Dell, HP, Lenovo, Microsoft Surface, MSI, with Acer and GIGABYTE to follow, arriving this fall (NVIDIA; Microsoft blog). "The PC is being reinvented," said Jensen Huang in NVIDIA's announcement. Industry coverage notes this positions NVIDIA into the consumer PC chip space while Microsoft optimizes Windows for the architecture (The Verge; Windows Blog).
According to NVIDIA's May 31, 2026 press release, NVIDIA unveiled RTX Spark, a new Arm-based superchip for Windows PCs that combines a 20-core Grace CPU, up to 6,144 Blackwell CUDA cores, and up to 128GB of unified memory, delivering roughly 1 petaflop of AI performance. NVIDIA said RTX Spark can run local large models up to 120B parameters with contexts up to 1 million tokens, render 90GB+ 3D scenes, and edit 12K 4:2:2 video (NVIDIA press release). MediaTek confirmed collaborative work on the chip's custom CPU design and is quoted in a PR Newswire release via The Manila Times (Vince Hu, MediaTek) describing the partnership (PR Newswire / Manila Times). PCMag and Japanese outlet PC Watch reported product-equivalence to NVIDIA's GB10/DGX Spark and listed OEM partners including ASUS, Dell, HP, Lenovo, Microsoft Surface and MSI (PCMag; PC Watch).
During a Computex keynote, Nvidia CEO Jensen Huang unveiled an Arm-based CPU called N1X and an integrated superchip family named RTX Spark, which CNBC reports will debut this fall in Windows laptops from Microsoft, Dell, HP, ASUS, Lenovo and MSI. Yahoo Finance reports the launch combines N1X and an Nvidia Blackwell GPU into a single package with 128GB of unified memory in the first-generation design. CNBC reports a Nvidia spokesperson said the company expects more than 30 laptops and 10 desktops to ship with the new chipline. CNBC also reports Nvidia said its Vera Rubin data-center AI chips are in full production with early customers including OpenAI, Anthropic and SpaceX. Seeking Alpha lists potential beneficiaries across the PC supply chain, naming memory, OEMs and storage vendors.
The U.S. Department of Commerce posted weekend guidance clarifying that export license requirements for advanced AI semiconductors apply to entities headquartered in China even when those entities are located outside China, the Reuters reporting by Karen Freifeld and Fanny Potkin summarized on May 31, 2026. The guidance names chips from U.S. producers including Nvidia's Blackwell and Rubin families and AMD's MI350x as covered items, and a Bureau of Industry and Security (BIS) spokesperson said, "BIS issued guidance clarifying export license requirements that have been in place since 2023," per Reuters. Reporting cites one chip industry source who estimated the scale of prior exports at "hundreds of thousands." The guidance does not force data centers to stop using already-deployed equipment, Reuters and CNBC report. Tech commentator Chris McGuire is quoted calling the loophole a "HUGE problem."
According to StockMarketWatch, Anthropic announced a $65 billion Series H funding round, producing a $965 billion post-money valuation. StockMarketWatch reports the round was led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, with participation from Blackstone, Brookfield, and Temasek. The coverage states the package includes $15 billion in previously committed hyperscaler investments, including $5 billion from Amazon, and that Anthropic has reached a reported $47 billion revenue run-rate. Separately, StockMarketWatch reports CBOE Global Markets received SEC approval to offer extended options trading beginning July 13, 2026, initially covering about 20 high-liquidity names and requiring a minimum $50 billion market cap to qualify.
Per NVIDIA's GTC Taipei announcement and accompanying press materials, NVIDIA introduced the NVIDIA Agent Toolkit, an open-source software stack for building long-running, autonomous enterprise AI agents. The Toolkit bundles open models, runtimes and domain skills including NemoClaw blueprints, Nemotron 3 Ultra models, the OpenShell runtime, and access to CUDA-X libraries, per NVIDIA. NVIDIA named enterprise partners and early adopters such as Adobe, Salesforce, SAP, ServiceNow, Cadence, Dassault Systèmes, Siemens, CrowdStrike and Palantir in coverage by NVIDIA and VentureBeat. Per NVIDIA, Nemotron 3 Ultra delivers 5x faster inference and up to 30% lower cost for complex agentic tasks. NVIDIA and multiple platform vendors including Microsoft, Canonical and Red Hat are integrating OpenShell and related tooling to offer secure runtimes across PCs, clouds and data centers, according to NVIDIA and partner announcements.
According to NVIDIA's press announcement on May 31, 2026, the Vera Rubin platform is ramping into full production to power "agentic AI factories" worldwide. NVIDIA's release states the platform unifies purpose-built racks including the Vera Rubin NVL72 systems, Vera CPU, BlueField-4 STX storage, and Spectrum-X photonics, and that hundreds of supply-chain partners across more than 350 factories in 30 countries are ramping production (NVIDIA Newsroom). The company claims the platform delivers 10x agent throughput versus the previous-generation Grace Blackwell platform, and the announcement includes a quote from Jensen Huang describing Vera Rubin as an "AI factory engine" (NVIDIA Newsroom). Independent coverage names major system builders and OEM partners in manufacturing and shipping Vera Rubin-based systems, including Dell Technologies, HPE, Lenovo and Supermicro (Wccftech; GlobeNewswire).
Nvidia has begun shipping its new Vera CPU systems to early customers, with first deliveries reported at Anthropic, OpenAI, SpaceXAI, and Oracle Cloud Infrastructure, according to PCMag, Datacenter.news and Yahoo Finance. PCMag reports Nvidia executive Ian Buck personally hand-delivered units to the four organisations. Nvidia and Datacenter.news report Vera uses 88 Olympus cores, offers up to 1.2 TB/s of memory bandwidth, and that per-core performance can be 50% faster under full load than prior designs. VentureBeat and Nvidia describe Vera Rubin as a seven-chip platform that pairs Vera CPUs with Rubin GPUs, and VentureBeat quoted Jensen Huang calling the platform a "generational leap." Yahoo Finance and PCMag report Oracle plans to deploy hundreds of thousands of Vera CPUs beginning in 2026.
OpenAI announced in a public post, according to Scientific American and Live Science, that an internal AI model produced a proof resolving the planar unit distance problem, a conjecture posed by Paul Erdős in 1946. Scientific American reports that mathematicians including Timothy Gowers and Daniel Litt reviewed the result and described the method as notable; Live Science says OpenAI released the prompt and described the model as a general-purpose reasoning model rather than a math-specific system. Singularity Hub and other coverage note follow-up work by human mathematicians and parallel results from other labs. Sources differ on technical details, but coverage frames this as the first AI-generated proof of this visibility and rigor, and as a milestone for mathematical AI research.
Anthropic announced the release of Claude Opus 4.8, available May 28, 2026, with benchmark and product updates, according to Anthropic's blog post and company materials. The company reports improvements versus Opus 4.7 across agentic coding, multidisciplinary reasoning, agentic computer use, and knowledge-work metrics, with sample gains such as agentic coding rising from 64.3% to 69.2% and a knowledge-work score increasing from 1753 to 1890 (Anthropic). Anthropic says Opus 4.8 fast mode is roughly 2.5x faster and the upgraded fast mode costs 3x less than before (Anthropic). The release also adds a research-preview feature called Dynamic workflows, an effort-control toggle in claude.ai and Cowork, and expanded Claude Code rate limits (Anthropic; 9to5Mac; TechCrunch). Coverage notes Anthropic compares Opus 4.8 to competitor models such as OpenAI's GPT-5.5 and Google's Gemini 3.1 Pro on select benchmarks (Yahoo Finance).
Per Anthropic's blog post (May 28, 2026), Claude Opus 4.8 (claude-opus-4-8) is available now and upgrades agentic coding, reasoning, and collaboration capabilities. Anthropic's announcement states fast mode runs at 2.5× speed and is three times cheaper than on previous Opus models, and the company reports usability features including user-selectable "effort" controls and a new Claude Code "dynamic workflows" capability. Independent testers and coverage report benchmark gains versus Opus 4.7 and competitive models on several agentic and coding tests, while third-party evaluations (Andon Labs) find regressions on some simulated economic benchmarks. Editorial analysis: practitioners should view the release as a pragmatic step toward more reliable agentic workflows, but also track failure modes that appear under long-horizon agentic stress tests.
According to the USC Viterbi School of Engineering, researchers built the Musician Hand, a four-fingered, tendon-driven robotic hand that self-explores a keyboard for two minutes and then reproduces an unheard melody after hearing it once. Per USC and reporting in TechXplore and Interesting Engineering, the system uses a short "motor babbling" phase to map finger actions to resulting sounds, analyzes audio with spectrograms and neural networks, and generates motor commands to play back notes and dynamics. Interesting Engineering reports the prototype reproduced a 30-note melody in a single attempt. The USC news release and allied coverage note a blind audition in which human judges sometimes could not distinguish the robot from human pianists. The USC release names Hesam Azadjou as lead author and Francisco Valero-Cuevas as corresponding author, and cites the paper titled "Perception in Action: A Robotic System that Can Teach Itself to Melodiously Play Music by Ear."
Researchers at the University of Osaka and collaborators introduced the Insect Synergy Circuit (ISC), a bio-hybrid AI system that monitors insect physiological signals to guide cyborg cockroaches, reporting simultaneous measurement of heartbeat, low-frequency neural features, and body motion (EurekAlert, TechXplore). The team built a lightweight wearable backpack for Madagascar hissing cockroaches that delivers low-burden stimulation via ultraviolet light and vibration, and uses machine learning to decide when to intervene (The Engineer, Interesting Engineering). Per the published study in ROBOMECH Journal, a Random Forest classifier trained on five conditions, baseline, ultraviolet exposure, chemical exposure, heat, and food, achieved 93% accuracy in classifying the insect's environment-associated internal state (Bioengineer, TechXplore). "The key shift is from 'controlling' to 'listening,'" said Professor Keisuke Morishima (EurekAlert).
Researchers from the University of Rochester, Brown University, and the University of Copenhagen used physics-informed AI together with MRI to estimate flow velocities in the brain's glymphatic system, according to a University of Rochester press release and coverage in Futurity and MedicalXpress. Their study, published in Science Advances, reconstructed fluid velocities from time-series MRI of dye spreading in brain tissue and animal experiments. The team reports two distinct transport regimes: a fast flow of a few microns per second around open regions such as the brain surface, and a much slower flow roughly 50x slower through deep tissue, per the University of Rochester materials. The work establishes baseline measurements in animals and reports the method could allow future comparisons with human imaging, according to the press release.
Per an AWS blog post, Anthropic's Claude Opus 4.8 is now available on Amazon Bedrock, with improvements aimed at agentic coding and long-running workflows. According to an Amazon announcement, the latest OpenAI models are available in limited preview on Amazon Bedrock, and the OpenAI coding agent Codex plus Amazon Bedrock Managed Agents are offered in limited preview as well. An Amazon press release describes CRED codelens running on Bedrock, reporting 4x AI acceleration and orchestration of 400+ specialized agents. The Bedrock product page states Bedrock powers generative AI for more than 100,000 organizations worldwide, and AWS Skill Builder lists a Bedrock course for developers. Editorial analysis: Cloud platforms combining frontier models, managed agents, and governance tooling lower integration friction for enterprises while raising expectations for production readiness.
Microsoft announced the Surface Laptop Ultra, a new 15-inch flagship Surface accelerated by NVIDIA's Arm-based RTX Spark, in company blog posts and press coverage (Microsoft Windows blogs; The Verge; Engadget). According to NVIDIA and Microsoft posts, RTX Spark delivers 1 petaflop of AI performance and can include up to 6,144 Blackwell RTX cores, up to 20 Arm CPU cores, and up to 128GB of unified memory (NVIDIA; Microsoft). Microsoft describes the Laptop Ultra as able to run models up to 120B parameters locally and highlights a 15-inch mini-LED PixelSense Ultra display with 2,000 nits peak HDR brightness (Microsoft Surface blog; The Verge). Reported availability is this fall; pricing has not been disclosed (Engadget; Windows Central). Industry context: this launch is a notable step for high-performance Windows-on-Arm hardware and local AI workflows.
A multi-country paper by Bick et al. (2026), reported in CEPR, the St. Louis Fed and Euronews, finds a clear gap in workplace AI uptake between the United States and Europe. The worker-and-firm surveys, covering more than 5,000 respondents in the US and six European countries plus a firm survey of 32 European countries, show higher company-level integration and individual use in the US: the paper reports 34% of US firms use AI for any purpose versus an EU average of 20%, and 43% of US workers report using AI at work versus 32% in Europe (Euronews summarising the research). The authors link much of the cross-country difference to management practices, with US workers more likely to report managerial encouragement and internal tooling, according to the study and reporting by CEPR and Euronews. Editorial analysis: this pattern echoes earlier ICT diffusion research and implies productivity trajectories will depend on how fast firms adopt AI-supporting management practices.
According to reporting by AwesomeAgents and StartupFortune, Visa made an undisclosed strategic investment in Replit and will integrate Visa Intelligent Commerce and the Trusted Agent Protocol into the Replit developer platform. AwesomeAgents reports that agents built on Replit can register in Visa's Trusted Agent Protocol registry and be verified at checkout, and that Visa Intelligent Commerce Connect reaches general availability in June 2026. StartupFortune and The New Stack note that more than 1,000 Visa employees already use Replit for prototyping. AwesomeAgents also cites Replit's recent fundraising: a $400M round in March 2026 at a $9B valuation and a target of $1B annualized revenue by year-end. Reporting frames the deal as embedding a payment identity layer directly inside the developer workflow for agentic applications.
According to NVIDIA, DLSS 4.5 Ray Reconstruction arrives in August 2026 and will be distributed through the NVIDIA App to all GeForce RTX GPUs, including RTX 20, RTX 30, RTX 40 and RTX 50 series (reported by NVIDIA and Videocardz). Videocardz and NVIDIA documentation state the update uses a second-generation transformer model with reportedly 35% more compute capability and processing 20% more parameters while maintaining similar runtime performance. NVIDIA and Videocardz list initial support for 27 games at launch and say Blender Cycles will integrate the denoiser in Blender 5.3 for viewport previews. Editorial analysis: The update represents an incremental step in shifting ray-traced denoising toward neural reconstruction, which lowers per-frame sampling requirements and standardizes denoising across engines.
Fortune reports Microsoft is developing a single Copilot "super app" that would combine GitHub Copilot, Copilot chat, Copilot Cowork, and a new agentic workflow capability internally named Autopilot. Fortune says the project is being spearheaded by Jacob Andreou, Microsoft's recently appointed head of Copilot, and that some elements could be referenced at Microsoft's event next week, though the app itself is not expected to be showcased. Fortune reports the company plans to launch the super app by the end of summer; the plans may evolve, the story adds. The report notes features under consideration include a toggle between personal and enterprise Microsoft 365 Copilots. Fortune says Microsoft declined to comment. The article places the move alongside other firms' attempts to unify AI tools under single interfaces.
SpaceX filed an S-1 for an IPO that public reporting frames as potentially the largest ever, with leaked and reported valuations ranging from about $1 trillion to $1.75 trillion (The Verge; Reuters) and estimates the offering could raise more than $50 billion (The New York Times) or roughly $75 billion (DW). The filings and reporting show heavy near-term losses driven by the companys AI unit, xAI, and disclose $4.3 billion in losses in the first three months of 2026, according to The New York Times. Reporting highlights Elon Musks tight control over governance, and outlets including NBC News and The Guardian describe large, performance-based compensation tied to ambitious Mars milestones. Industry coverage emphasizes investor FOMO, underwriting complexity, and the unusual mix of space infrastructure and AI in the valuation case.
CNBC reports that Dell Technologies posted 88% year-over-year revenue growth for the quarter and adjusted EPS of $4.86, beating LSEG consensus. Revenue was $43.84 billion, CNBC says, and the stock rose as much as 19% in extended trading. CNBC attributes the surge to demand for AI-capable servers and networking equipment, reporting that AI server revenue grew 757% year over year to $16.1 billion. CNBC also reports Dell has raised its full-year AI revenue outlook to $60 billion, up from a February projection of $50 billion, and that the company raised prices in January amid a global memory shortage. CNBC additionally notes a Pentagon award worth $9.7 billion and that the stock is up more than 150% year to date.
The New York Times reports two super PACs allied with major AI firms are locked in a bitter feud during the 2026 midterms. One group, Public First, is allied with Anthropic and the other, Leading the Future, is aligned with OpenAI, the Times notes. The New York Times reports the groups have already laid out nearly $24 million and that supporters have pledged more than $100 million in additional spending. The Times quotes Democratic strategist Cooper Teboe calling the conflict "a war for the future." Commstrader reports Public First's leader Brad Carson described the rivalry as "matter and antimatter," and describes tactical clashes including canceled ad buys and pressure on candidates.
Reporting by Bloomberg states that investors are increasingly targeting Asian supply-chain companies after an anticipated wave of US stock offerings by SpaceX, Anthropic PBC and OpenAI. Bloomberg reports the thesis: the "billions of dollars" these companies are set to raise will trigger fresh technology spending, with capital likely flowing to makers of server parts, specialized materials, cooling components and power equipment. Bloomberg frames those suppliers as potential beneficiaries that could help drive a new leg of the rally in Asian stock markets. The coverage highlights investor positioning rather than confirmed corporate procurement plans.
Amazon and Uber are pulling back from indiscriminate AI consumption after internal leaderboards and runaway bills exposed costs. According to the Financial Times and reporting cited by TheStreet and Business Insider, Amazon deprecated an employee-built token-usage leaderboard called KiroRank; Business Insider reports Dave Treadwell, Amazon senior vice president of engineering, told staff, "Please don't use AI just for the sake of using AI." Fortune and Investing.com report Uber exhausted its 2026 token budget in the first four months, driven in part by heavy use of Anthropic's Claude Code. The Wall Street Journal reports executives across industries are starting to ration AI as compute and model-access costs rise. Broader coverage in Fortune, CNET, and Yahoo Finance frames this as a retreat from 'tokenmaxxing' toward tighter cost and ROI discipline.
Exponential View publishes an essay arguing that widely adopted developer tools and generative AI are raising individual productivity without producing proportional firm-level returns. A senior executive at a "well-known public tech company" told Exponential View that about a thousand engineers using Claude Code are producing more code and pull requests, yet organisational gains are smaller: "one plus one plus one plus one equals one-and-a-half," the executive said. Exponential View also quotes Uber COO Andrew Macdonald observing that it is "very hard to draw a line" between AI outputs and a clear increase in consumer-facing feature value. The essay frames the gap using the economic literature on general-purpose technologies, citing Robert Solow and Paul David to explain complementary investments and adoption lags.
At GTC Taipei / COMPUTEX, Nvidia founder and CEO Jensen Huang used a keynote to highlight agentic and physical AI, new hardware, and partner agreements. According to NVIDIA's corporate blog, Huang said, "AI is now a profit generator. AI is now a GDP generator." The Wall Street Journal reports Nvidia introduced the first personal PCs designed for running AI agents and will work with manufacturers including Dell, Lenovo, HP, Asus, and MSI. Seeking Alpha and Nvidia coverage note a new Vera CPU (reported as delivering 1.8x faster task completion by Seeking Alpha), an RTX Spark superchip (WSJ), and collaborations with software and systems firms including Cadence, Siemens, Synopsys, CrowdStrike, and Palantir (Seeking Alpha). Yahoo Finance and other reporting cite partnerships with Foxconn and Taiwan medical centers to develop agentic physical-AI robotics in healthcare.
Mukesh Ambani announced at the India AI Impact Summit that Reliance Industries and its digital arm Jio will invest Rs 10 lakh crore over seven years to build multi-gigawatt-scale, AI-ready data centres, according to DD News and ANI. RIL's Integrated Annual Report 2025-26 and a post on X reiterate the commitment and describe it as part of building "sovereign AI infrastructure," per ANI and ril.com. Economic Times reports construction has begun at Jamnagar, with over 120 MW expected online in the second half of 2026 and a stated pathway to gigawatt-scale compute. The ET coverage also says the deployment will pair edge compute tightly with Jio's network and use up to 10 GW of green-power surplus anchored by solar in Kutch and Andhra Pradesh.
At Computex in Taipei, Nvidia CEO Jensen Huang showcased new laptop system-on-chips, the N1 and N1X, and promoted the RTX Spark platform, according to Gizmodo. Gizmodo quotes Huang saying RTX Spark will handle "every application that Windows has ever run." Leaked specifications reported by Tom's Hardware list a top-end N1X with up to 20 Arm CPU cores, 6,144 CUDA cores, and support for up to 128GB of LPDDR5X memory, while the N1 is reported in 10- and 12-core configs and up to 64GB of memory. The chips pair Arm CPUs with Nvidia Blackwell GPUs and are tied to Windows on Arm interest; The Verge and Windows Central report coordinated social posts from Nvidia, Microsoft, and Arm teasing the announcement. Editorial analysis: For practitioners, the launch underlines the industry shift toward on-device and low-latency AI acceleration, but software compatibility and x86 emulation remain open technical and adoption questions.
Tom's Hardware reports that at Computex Intel revealed details for the next-generation data-center GPU codenamed Crescent Island, built on the Xe3P architecture and described as "built for agentic AI." Tom's Hardware reports the Crescent Island reference design is a PCI Express add-in card with a 350W power target and uses LPDDR5X instead of GDDR or HBM. Tom's Hardware reports the reference includes 160GB of LPDDR5X and that the design allows partners to scale to 480GB. The article notes leaked analysis suggesting a wide, slow 640-bit bus and calculates about 684 GB/s of bandwidth using 10.7 Gbps LPDDR5X parts; Tom's Hardware also notes 24GB LPDDR5X modules are available from vendors such as Samsung.
According to Daniel Sparks at The Motley Fool (via Yahoo Finance), Nvidia CEO Jensen Huang said on the company's fiscal first-quarter earnings call, "Demand has gone parabolic." Motley Fool reports Nvidia's fiscal first-quarter 2027 revenue was $81.6 billion, up 85% year over year; its AI data center revenue was $75.2 billion, up 92% year over year; and earnings per share were $1.87, up 140% year over year. Motley Fool also notes Nvidia increased capital returns and raised its quarterly dividend following the quarter. The article frames Huang's comment and the results as reinforcing a strong AI-driven demand environment for GPUs and related infrastructure.
Per PR Newswire and NVIDIA, Unitree Robotics and NVIDIA announced the H2 Plus reference humanoid robot built on the Isaac GR00T platform, combining a Unitree H2 humanoid body, Sharpa Wave five-finger hands, and onboard Jetson Thor compute (NVIDIA press release). NVIDIA named leading research institutions including Ai2, ETH Zurich, Stanford Robotics Center and UC San Diego as early users (NVIDIA). Unitree founder Xingxing Wang and NVIDIA CEO Jensen Huang provided quoted remarks in the launch materials (PR Newswire; NVIDIA). CNBC reports sales primarily to research institutions are set to begin later this year (CNBC). Editorial analysis: This reference-design announcement reduces integration overhead for research teams by packaging body, hands, on-device Blackwell-class compute, and an open software stack into one validated starting point.
A popular npm package, codexui-android, secretly exfiltrated OpenAI Codex authentication tokens, researchers report. According to Aikido Security researcher Charlie Eriksen, the package amassed about 27,000 weekly downloads and, starting roughly a month after publication, every invocation began sending the contents of users' auth.json (Codex auth tokens) to an attacker-controlled endpoint, reportedly sentry.anyclaw.store (per Aikido blog). Cybernews reports the package delivered genuine functionality and that the malicious code was pulled at runtime rather than present in the GitHub repository, which allowed it to evade source audits and, Cybernews adds, to bypass Google Play pre-publication scans. The incident illustrates a supply-chain technique where threat actors weaponize legitimate developer tools, according to the published researcher findings.
The piece demonstrates methods to break Google's reCAPTCHA, framing captchas as mechanisms that distinguish human users from automated programs. It notes many internet service providers have a major incentive related to captchas.
According to a report by the Financial Times, Iranian-linked operators have used Western AI tools including ChatGPT and Gemini to develop malware, craft phishing campaigns in Hebrew and Arabic, and build fake online personas targeting the US and Israel. India Today, summarizing the Financial Times coverage, reports that operators used AI-generated identities to build trust with victims and that Israel experienced waves of phishing emails and text messages linked to these campaigns. Per the Financial Times as reported, Google detected the state-linked group APT42 using Gemini in late February. India Today also reports that the United Arab Emirates said in May it faced more than more than half a million cyberattacks daily, some reportedly assisted by AI tools. Editorial analysis: Industry practitioners should treat this reporting as another data point in a growing pattern where accessible large language models are leveraged for social-engineering and low-cost automation of malicious tradecraft.
Palo Alto Networks published a security advisory for an authentication-bypass vulnerability in PAN-OS GlobalProtect, tracked as CVE-2026-0257, that the CVE record lists with a CVSS 7.8 score, according to CVE.org. Multiple vendors and researchers, including Rapid7 and BleepingComputer, report that the flaw is being actively exploited in the wild. SentinelOne and Rapid7 technical summaries describe the bug as an authentication-state validation weakness in the GlobalProtect portal and gateway that can let unauthenticated attackers establish VPN sessions. Reporting indexed from CISO Series and Cybersecurity Headlines says attackers have abused ChatGPT "share" links to host fake outage pages that deliver malware. Reporting by CISO Series also highlights a federal audit that found problems in NIST's National Vulnerability Database (NVD). Editorial analysis: companies relying on enterprise VPNs should treat active exploitation of authentication-bypass vulnerabilities as high operational risk and prioritise visibility and patching in their vulnerability management workflows.
A correspondence in The Lancet by Maxim Topaz and colleagues reports an audit of 2.5 million biomedical papers in PubMed Central Open Access covering January 1, 2023 to February 18, 2026. The authors identified 4,046 fabricated references across 2,810 papers from a set of 97.1 million verified references, per the correspondence (Topaz et al., The Lancet, May 9, 2026). The research team developed an automated verification pipeline and used an LLM review step, reported by Forbes and Columbia University, including Claude 3.5 Haiku to triage flagged items. The Lancet correspondence and Columbia press materials link the sharp rise in fabricated citations since mid-2024 to the increasing use of AI writing tools, and the authors recommend publishers verify references and indexing services add metadata to track fake citations.
Per an Anthropic engineering blog post, the company explains two primary approaches to limiting agent risk when deploying Claude across products: human-in-the-loop supervision and containment via access boundaries such as sandboxes, virtual machines, and egress controls. The post reports telemetry showing users approved roughly 93% of permission prompts under a prior Claude Code gating flow and describes the introduction of Claude Code auto mode to reduce approval fatigue. Anthropic also states that Claude Mythos Preview was judged to have a blast radius too high to ship in April 2026. The post frames containment engineering as the main focus for reducing the potential damage of increasingly capable agents while acknowledging that risk cannot be reduced to zero.
Per the Vatican text published on the Holy See website, Pope Leo XIV released an encyclical titled Magnifica Humanitas: On Safeguarding the Human Person in the Time of Artificial Intelligence (signed May 15; published May 25). The 42,300-word document addresses the social, moral, environmental, and economic effects of AI and calls for stronger regulatory tools, warning that AI "must be disarmed" and should not be allowed to "dominate humanity" (quoted in the encyclical; reported by The Atlantic and PBS). Major outlets including The New York Times, Time, The Washington Post, and NPR covered the papal text and its implications for Big Tech, labor, and international affairs. Editorial analysis: Industry observers will treat the encyclical as a high-profile moral intervention likely to amplify public and political pressure around AI governance rather than a direct regulatory blueprint.
China tightened rules on overseas investments after blocking the Meta-Manus deal. The new controls affect foreign investments in the technology and national security sectors.
Reporting by the Associated Press shows the Trump administration is pressing to expand use of artificial intelligence in the U.S. military while some senior officers and tech companies urge safeguards. Adm. Frank Bradley, head of U.S. Special Operations Command, told a special forces conference in Tampa that "we have to be very careful about how we come to (AI's) employment and its inspiration into the delivery of lethality," and warned humans must have confidence AI will "deliver violence only where we intend it to be delivered" (AP). The AP coverage says Defense Secretary Pete Hegseth has pushed for rapid AI integration and told an audience he would reject models "that won't allow you to fight wars." AP reporting adds a Pentagon official said the department is focused on creating "functional battlefield tools." Some outlets report clashes with companies, including Anthropic, over safety concerns.
El Pais reports that academics Eliezer Yudkowsky and Nate Soares argue in their new book, If Anyone Builds It, Everyone Dies, that continued improvements in artificial intelligence could lead to human extinction, possibly "in a matter of months or within a decade," according to the article. El Pais also notes the authors were signatories to a wider 2023 open letter calling for a six-month research moratorium, and cites a past Yudkowsky piece that recommended extreme enforcement measures including having their data centers "destroyed by air strike." The article contrasts this catastrophist view with accelerationists, who argue advanced AI could cure diseases and increase prosperity. Editorial analysis: The El Pais piece frames the debate as a central fault line in AI policy and ethics, underscoring why regulators and practitioners continue to weigh safety, governance, and fast technical progress.
According to dimsumdaily.hk, Demis Hassabis, co‑founder and chief executive of Google DeepMind, warned that artificial intelligence is a "species-level transition" with "little margin for error" over the next decade and urged coordinated international regulation within five to 10 years. At a Stanford GSB event, Hassabis said AI stands in "the foothills of the singularity" and is advancing roughly 10 times faster than the Industrial Revolution. He likened governance to nuclear non-proliferation and climate change, called frontier systems profoundly dual-use, and raised concerns about open-source releases enabling "bad actors." Per the report, he backed "smart, targeted" oversight such as periodic independent evaluations of advanced models and sector-specific rules for driving and medicine, and defended DeepMind's decision to release AlphaFold predictions citing open crystallography traditions.
President Lee Jae Myung pursued aggressive, AI-focused policy pushes during his first year in office aimed at elevating Korea's standing; the government effort has laid groundwork for businesses, while tangible outcomes have yet to materialize.
Team Global Express has identified 73 AI use cases and put 12 "AI agents" into production, with a further five proof-of-concepts underway, Michael Farrar told the AWS Summit Sydney, as reported by ITNews. Farrar said the company consolidated previously siloed systems into a centralised data platform on AWS, moving to Redshift for analytics and DynamoDB for operational stores, and has been adding metadata and guardrails to improve data hygiene. In production, the company uses Amazon Rekognition to strip personally identifiable information from proof-of-delivery images and is building frontline operational intelligence with Amazon Bedrock AgentCore, per the ITNews report. Separately, a Salesforce customer case study and Agentforce Summit coverage describe an AI agent named Āwhina that Team Global Express deployed into track-and-trace workflows, with Salesforce reporting call-volume reductions of 25% in weeks and an estimated 50% reduction overall.
Researchers at Washington University School of Medicine developed an AI classifier that distinguishes between four common causes of dementia-Alzheimer's disease, Parkinson's disease, frontotemporal dementia, and dementia with Lewy bodies-as well as healthy brain aging, reporting over 90% accuracy (WashU Medicine; Futurity). The team selected a panel of 15 proteins measured in blood and trained and tested the model on blood-protein data from more than 3,200 individuals, per Futurity. The classifier can also detect coexisting disease processes, a frequent clinical challenge. The findings appear in the journal Alzheimer's & Dementia, and senior author Carlos Cruchaga is quoted on the limits of single-diagnosis approaches (WashU Medicine).
A randomized clinical trial published in JAMA Network Open found that a conversational AI platform, Kai, produced modest but statistically significant reductions in self-reported anxiety and depression among about 1,000 Israeli university students, per the publication. Per the trial, after 12 weeks the AI group had better GAD-7 scores than face-to-face group therapy (mean difference -2.17 points, 95% CI -2.67 to -1.67) and controls (mean difference -2.15, 95% CI -2.65 to -1.65), and better PHQ-9 scores versus controls (mean difference -1.99, 95% CI -2.63 to -1.35) (MedPage Today; JAMA Network Open). MedPage Today and other reporting note absolute mean GAD-7 fell from 7.27 to 6.28 and PHQ-9 from 7.64 to 6.68 in the AI arm, changes below commonly cited minimally clinically important differences, and that no serious adverse events were reported. Editorial analysis: The randomized design and sample size strengthen credibility, but effect sizes were small, suggesting conversational agents are likely to function as scalable adjuncts or early-intervention resources rather than replacements for established therapies.
Business Insider reports that generative AI is making polished, hyper-personalized cover letters trivially easy to produce, undermining a longstanding signal recruiters used to find standout applicants. The article quotes Wharton professor Judd Kessler saying he can no longer distinguish strong candidates from cover-letter text alone: "I used to get really good cover letters, and be like, 'oh, I should really talk to this person, and prioritize those people,'" Kessler says, per Business Insider. Business Insider also notes the story is subscriber-only. The piece frames the trend as creating a noise problem for hiring: many applicants now appear "exceptional" on paper, complicating early screening for admissions and hiring teams.
The Korea Advanced Institute of Science and Technology (KAIST) held an "AI College Vision Declaration Ceremony" on June 1, according to Seoul Economic Daily. Seoul Economic Daily reports the new KAIST AI College began operating an undergraduate program this spring semester and plans to offer more than 50 AI-specialized courses aimed at connecting problem definition to AI modeling using real industrial data. The ceremony included a keynote from Bae Kyung-hoon, Deputy Prime Minister and Minister of Science and ICT, who said, "Cross interdisciplinary boundaries to design your own AI major, and create AI that pursues humanity by solving the challenges of various industrial sites." Seoul Economic Daily also reports KAIST will recruit private-sector experts as adjunct faculty and that Dean Yoon Kuk-jin announced plans to build a "full-stack" AI education and research system integrating AI core technology, systems and infrastructure, and AI+X convergence.
AP reports that a wheeled caregiver robot nicknamed Robbie, officially model Stretch 4, is being piloted in the Durham, New Hampshire, home of Brenda and Brian Marquis to provide reminders, guided exercise and prompts to eat and drink. The University of New Hampshire laboratory operates the device with funding from the National Institute on Aging, per AP coverage republished by ABC News and Journal-News. The robot uses a small digital-screen "face" to play exercise videos and asks spoken prompts such as "Do you want to exercise now? Please answer yes or no," which is shown in AP reporting. Editorial analysis: Pilot deployments like this illustrate practical, narrow uses for assistive robots in household routine support, but the story is a single-site demonstration rather than evidence of broad commercial readiness.
