1. Anthropic expanded Project Glasswing to ~150 new organizations across 15+ countries, covering power, water, healthcare, communications, and hardware sectors.
2. The company confidentially filed for an IPO on Monday, following a $65 billion funding round at a nearly $1 trillion valuation.
3. Claude Mythos, Anthropic’s most powerful model, identifies thousands of zero-day vulnerabilities over several weeks.
4. Initial 50 partners, including the U.S. government, gained access to Mythos Preview in early April.
5. A successful attack on partner codebases could affect over 100 million people with global and national security ramifications.
6. New partner countries include Australia, Canada, France, Germany, Italy, Switzerland, Netherlands, Spain, Belgium, Sweden, India, Japan, New Zealand, and South Korea.
7. Named organizations receiving access include Okta, Samsung, SK Hynix, SK Telecom, NATO, and ENISA.
8. Anthropic expects other AI firms to develop comparable models soon, driving rapid safeguards within Project Glasswing.
9. OpenAI released its own cybersecurity model, GPT-5.5-Cyber, to a large group of partners for testing.

1. Nvidia unveiled the RTX Spark superchip at Computex, a 1-petaflop PC CPU designed to run AI agents like OpenClaw or Hermes Agent securely.
2. RTX Spark Windows PCs will be available this fall from ASUS, Dell, HP, Lenovo, Microsoft Surface, MSI, with Acer and Gigabyte to follow.
3. The PCs include secure sandboxes jointly developed with Microsoft, and enough CPU, GPU, RAM, and Nvidia CUDA software to run local LLMs.
4. Nvidia stated RTX technology delivers faster AI performance, better image quality, and supports AI features in over 1,000 games and applications.
5. Over 100 Windows software makers have signed on to support RTX Spark, including Adobe, Blender, ComfyUI, Riot Games, and Xbox.
6. Jensen Huang’s vision: with RTX Spark and Windows, users ask and the PC does the work, eliminating launching apps, pointing, clicking, and typing.
7. Last month (May 2026), Huang promised a new $200 billion market for Nvidia in selling CPUs for AI, citing $20 billion already sold of the Vera server CPU.
8. Nvidia ARM-based Windows devices previously failed in 2013 with Microsoft writing off $900 million on Surface RT, but Huang's current PC pursuit is backed by record quarterly revenue.
9. Microsoft named its RTX Spark PC the Surface Laptop Ultra, calling it “the most powerful Surface Laptop ever built.”
10. PC manufacturers have not released pricing; systems appear to be full-fledged Windows versions of the DGX Spark mini-computer sold for about $4,800 to developers.

1. WWDC 2026 begins Monday at 10 a.m. PT/1 p.m. ET, streamed via Apple Developer app, website, and YouTube.
2. Siri will receive a major AI upgrade leveraging Google's Gemini technology for conversational context and multi-step tasks.
3. Leaks indicate a standalone Siri app to compete with ChatGPT, Claude, and Gemini, with optional auto-delete conversation timers (30 days, 1 year, or indefinite).
4. Apple plans an AI agent integration with the App Store, enabling task delegation like booking, document editing, and smart home control.
5. The Camera app will introduce a "Visual Intelligence" section with a dedicated Siri mode, using Google Image Search for object identification.
6. The Photos app will get AI-powered scene recommendations, automatic object removal, and natural language editing requests.
7. The Wallet app will add a bill-splitting feature that photographs receipts and generates payment requests, plus a "Create a Pass" option for digitizing physical items.

1. SoftBank Group plans to invest 45 billion euros ($53 billion) over five years for AI infrastructure in France, as part of a 75-billion-euro program to roll out 5 GW of AI data center capacity.
2. The initial phase involves building 3.1 GW of AI data centers in northern Hauts-de-France by 2031, including Dunkirk, Bosquel, and Bouchain.
3. This commitment marks SoftBank’s largest AI infrastructure investment in Europe.
4. SoftBank overtook Toyota to become Japan’s most valuable company by market cap, according to FactSet data.
5. CEO Masayoshi Son confirmed the 75 billion euro investment and signaled the full system cost approaches $750 billion, with project financing aggregated and hyperscalers as customers.
6. SoftBank partners with Schneider Electric to develop a large-scale industrial production cluster in Dunkirk.
7. SoftBank shares rose 14% on Monday and are up over 70% year-to-date in 2026.

1. DeepSeek is raising ~50 billion yuan ($7 billion) in its first external funding round, with a post-money valuation of 350–400 billion yuan ($49–56 billion).
2. Founder Liang Wenfeng is contributing 20 billion yuan ($3 billion) of his own capital.
3. Tencent plans to invest 10 billion yuan ($1.5 billion) and CATL commits 5 billion yuan ($740 million) as the largest external investors.
4. DeepSeek is in final investment talks with China’s National AI Industry Investment Fund, NetEase, and JD.com, with fewer than 10 investors expected to participate.

1. Alphabet plans to raise $80 billion via stock sale, with $10 billion sold to Berkshire Hathaway, to fund AI infrastructure and global compute.
2. Alphabet cites strong enterprise and consumer demand for AI solutions exceeding available supply as the reason for scaling investments.
3. Sundar Pichai stated at Google I/O last month that Alphabet expects $180 billion to $190 billion in capex this year.
4. Google and other tech giants are expected to spend as much as $700 billion on AI capex this year.

1. Anthropic confidentially filed with the SEC to kick off its IPO process, setting the stage for a massive public offering.
2. As of its last fundraise, Anthropic holds a post-money valuation of $965 billion, making it the world’s most valuable startup, exceeding OpenAI’s $852 billion valuation.
3. SpaceX plans to IPO on June 12, 2026, raising $80 billion, which would be the largest IPO in history.
4. SpaceX owns xAI and has a deal with Anthropic paying $15 billion per year for use of SpaceX data centers.
5. OpenAI’s legal battle with Elon Musk ended with a dismissal of all claims due to the statute of limitations.

1. Xpeng began series production of its first Robotaxi model in its Guangzhou plant, announced in November 2025 as one of three models for 2026.
2. The Robotaxi is based on the six-seat GX SUV (5.27 meters long, revealed February 2026) and uses Level 4 autonomous driving standards.
3. It is powered by four self-developed Turing AI chips delivering 3,000 TOPS onboard computing power, avoiding cloud dependency for real-time road calculations.
4. Xpeng follows Tesla’s camera-only approach, using the VLA 2.0 (Vision-Language-Action) model to eliminate a language translation step and reduce system reaction time to under 80 milliseconds.
5. VLA 2.0 is also deployed in the ID.UNYX 08 from VW’s China partnership with Xpeng.
6. Xpeng established a dedicated Robotaxi business unit in March 2026 to handle product definition, R&D, and operations for commercialization.
7. Amap (Alibaba’s mobility app) is named as the first global ecosystem partner for the Robotaxi SDK, suggesting Xpeng will operate Robotaxis and make them bookable via ride-hailing platforms.
8. Xpeng has been testing autonomous vehicles on public roads in Guangzhou since January 2026 and plans pilot Robotaxi projects in the second half of 2026.
9. Xpeng targets fully autonomous operations without safety personnel by early 2027.

1. MB. Drive Assist Pro is an SAE Level 2 system requiring drivers to stay attentive (hands-off, eyes-on) and is currently limited to inner-city driving.
2. Mercedes development chief Jörg Burzer announced the system will be available in individual German cities by end of 2026 and nationwide by early 2027, starting with Stuttgart and Munich.
3. A China-specific version has been available since autumn 2025 in partnership with Momenta, while the western version uses Nvidia and was unveiled at CES in January 2026.
4. The system merges driver assistance and navigation, allowing the vehicle to navigate from parking spot to destination with cooperative steering that does not require deactivation.
5. Mercedes CEO Ola Källenius reported a test drive in San Francisco exceeding one hour without any manual intervention.
6. MB. Drive Assist Pro replaces the SAE Level 3 Drive Pilot, offering urban autonomous driving but with driver supervision, while Level 3 was limited to highway scenarios at up to 95 km/h.
7. The Level 2 system avoids costly laser-based lidar sensors, enabling significantly lower cost and broader everyday usability compared to Level 3.

1. Experts warn that China is catching up to the U.S. on autonomous driving due to market conditions, regulatory changes, and faster vehicle development.
2. Tesla’s Full Self-Driving Supervised launched in China with an update following years of delays, but approval for fleet-wide push remains pending.
3. Waymo is poised to compete against Baidu in London, and Chinese brands expand into global markets including North America.
4. Chris Liu of Omdia likened China’s catch-up in autonomous driving to its earlier rapid progress in battery technology through R&D and scaling.
5. Tesla’s hands-free FSD is considered a global leader in AI-based automated driving, while competitors Xpeng and Xiaomi advanced semi-autonomous features in China.
6. Hyper-competition (involution) in China’s EV market drives companies to differentiate via ADAS, but BYD undercuts subscription pricing, per Tu Le.
7. Chinese government policy guidance on price has eased pressure, but brands like BYD make advanced driving tech standard on many models.
8. China introduced restrictions on Level 2 hands-off driving (warnings after 1 minute) at end of last year, unlike the U.S. permit approach.
9. Jeremy Carlson noted China’s “ask for permission, then forgiveness” approach vs. U.S. “do what you like and be prepared for consequences.”
10. Many vehicles in China can drive parking spot to parking spot for around $30,000, providing scale of data to speed up learning.
11. China’s MIIT started issuing Level 3 production and testing permits; some U.S. states and Japan allow Level 3, with GM and Ford planning Level 3 in 2028.
12. Congress pushes NHTSA for a consistent national AV policy to replace state patchwork.
13. Stellantis announced three partnerships (including Wayve) to launch hands-free door-to-door supervised autonomous driving in North America in 2028.
14. Waymo provides over 500,000 trips per week across 11 U.S. metro areas and expands internationally.
15. China’s engineering workforce size, tech-savvy consumer base, EV strengths, and faster vehicle development (2–3 years ahead) offer advantages.
16. Data regulations in Europe force Chinese firms to collaborate with foreign suppliers for ADAS, leading to polarized regions.
17. Tu Le questioned whether U.S. systems can be called global if Tesla cannot launch full FSD in China and current state holds.

1. Uber invested nearly $500 million in Nuro through direct investments and milestone-linked funding, including a $203 million round valuing Nuro at $6 billion.
2. Nuro, backed by Nvidia and SoftBank, has a three-way agreement with Uber and Lucid to deploy 35,000 robotaxis using Lucid Gravity SUVs and midsize models with Nuro’s software on Uber’s network.
3. Uber’s follow-on investment in Nuro was considerably larger than its initial outlay, with additional funds contingent on hitting defined development and commercial milestones.
4. The first set of milestones has been met on schedule, unlocking an initial tranche of performance-linked capital, with outstanding payments tied to driverless testing in 2026, autonomous passenger operations before year-end, and service expansion in 2027.
5. Nuro pivoted in 2024 from small delivery robots to licensing its autonomous driving software to vehicle manufacturers and mobility operators.
6. In April 2026, Nuro obtained a California permit to test autonomous Lucid Gravity vehicles without a safety driver, and in May 2026, authorization to carry passengers in supervised testing.
7. Uber has additional autonomous vehicle partnerships with Baidu, Rivian, Wayve, and collaborates with Waymo in selected US cities.

1. Tencent is advancing plans to launch an embedded AI agent within WeChat, testing a prototype that can complete tasks inside the app.
2. Tencent plans to begin the regulatory approval process required before a public rollout as early as June 2026.
3. After regulatory approval, Tencent intends to test with a small group of external users before gradually expanding availability.
4. Full public launch timing remains uncertain due to unclear regulatory review duration.
5. Users can access the AI agent’s chat interface by swiping right from the WeChat home screen.
6. The agent can automatically connect with WeChat’s millions of mini programs to execute commands like finding cafés by taste and price preferences and ordering drinks.

1. A humanoid robot from X Square Robot spent three hours at Daniel Wang's Beijing apartment, folding about three pieces of clothing and arranging shoes, with most chores done by a human housekeeper, for a service fee of 149 yuan ($22).
2. The primary purpose of the robot was to collect training data from a real household, with the homeowner consenting to privacy exposure.
3. Robotics development globally is constrained by a shortage of training data combining complex visual and movement information; teleoperation is costly, time-consuming, and fails to generalize to real-world environments.
4. Chinese tech companies are leveraging low labor costs, government support, and public enthusiasm to mobilize large populations for robotics data collection in realistic settings.
5. Since the beginning of 2026, developers globally have ramped up collecting egocentric data (first-person videos of human hands performing tasks) in households, retail shops, and factories.
6. JD.com is working with the Suqian government to generate 10 million hours of robotics training data over two years via a dedicated "data collection neighborhood" where residents film themselves doing chores.
7. JD.com plans to eventually bring 100,000 employees and 500,000 external workers into its data collection operations.
8. Two data vendors in Guangdong are working with dozens of electronics and packaging factories, having workers wear head cameras and wrist sensors to record hand movements.
9. Some factory owners are reluctant due to potential worker slowdowns, as reported by data supplier Carson Lin in Dongguan.
10. According to analyst Marco Wang (Interact Analysis), China leads in hardware and data ecosystem for robotics, while the U.S. leads in top-tier AI talent and model research.
11. Oregon State University professor Alan Fern notes that applying the scaling logic of large language models to robotics via teleoperation data and egocentric videos is unproven.
12. A stay-at-home mom, Gao Bo, films herself doing chores for six hours per day at 20 yuan ($3) per hour, creating a rare job that allows care for her teenage son.

1. The EU executive proposes “technological sovereignty” to eliminate foreign “kill switch” access to vital tech services, reducing dependencies on the US and China.
2. European Commission on Wednesday said the bloc must cut foreign supplier reliance in cloud computing, AI, and semiconductor production.
3. Vulnerabilities were exposed last year when China stopped semiconductor exports, nearly halting the European car industry.
4. Henna Virkkunen stated the 2018 US Cloud Act allowing US federal data access in other countries is not aligned with EU rules.
5. Virkkunen said the EU will not work in isolation but must identify “risky dependencies” and ensure sensitive security/law enforcement data is controlled in Europe.
6. The EU requires that no foreign entity holds a “kill switch possibility”; US cloud providers face “very difficult” criteria to serve EU defence.
7. The EU relies on foreign providers for over 80% of digital products, services, infrastructure, and intellectual property.
8. Proposals require EU member states to conduct risk assessments on cloud providers in defence, criminal justice, and border management, potentially switching providers.
9. US cloud providers may have to comply with EU data protection rules and prove no compelled surrender of EU data to US authorities.
10. The Computer and Communications Industry Association (Amazon, Google) called the proposals a “dangerous recipe for progressive market shutdown.”
11. The Commission aims to boost EU AI infrastructure by fast-tracking datacentre builds and promoting European semiconductor production.
12. Olivier Darmouni of HEC Paris doubted the realism of building an EU “advanced manufacturing facility” for cutting-edge AI chips at needed timescale, but building RAM and memory chips could be beneficial.
13. The EU produces only 10% of world semiconductors and is almost entirely dependent for advanced AI chips; building fabrication plants is multi-year, multi-billion-dollar.
14. The plan to triple EU datacentre capacity within five to seven years includes a rating system for energy efficiency and “acceleration zones” for fast-tracked permitting.
15. Darmouni said the proposals dodge reconciling datacentre buildout with climate goals, risking electricity bill hikes and climate trajectory damage.

1. On June 1, 2026, Salesforce announced a $2 billion investment in France through 2030, building on a previous $3.5 billion commitment, including a new AI Innovation Hub in Paris.
2. The Paris AI Innovation Hub is Salesforce's first in the EU, following its London AI Center, for co-innovation, testing, and deployment of agentic AI.
3. Salesforce is hiring in cybersecurity, data, agentic AI, and deployment engineering in Paris and region to support French customers' agentic transformation.
4. Salesforce enables French customers to control encryption keys via Thales or Eviden, provides real-time activity transparency, supports local LLM integration with Mistral, and deploys MuleSoft, Tableau, and Informatica on sovereign cloud platforms (S3NS, Scaleway, Cloud Temple, Orange).
5. Bouygues deployed Agentforce-powered agent IRIS to 12,000+ employees, delivering 95% accuracy.
6. Laboratoires Pierre Fabre deployed "One CRM" to 4,000+ users to unify pharma and dermo-cosmetics activities.
7. The Adecco Group completed 228k agent conversations globally, with 162k pre-screening conversations at 87% response rate, powering 30,000 monthly interactions, and enabling 40% faster qualified CV sharing.
8. Salesforce Ventures invested $350 million+ in the region, backing Mistral, Hugging Face, and OpenClassrooms.
9. Salesforce contributed $16.3 million in grants and 224,000 volunteer hours in France, plus new $2.6 million in grants for career readiness and digital skills.
10. Salesforce launched an Agentic Academy with Simplon Foundation to train at least 2,000 people across 80 organizations, supported by 160 employees volunteering 1,000+ hours.

1. METR's updated research shows AI task-completion capability is now doubling every 130 days (approximately 4.3 months), accelerating from the prior estimate of every seven months.
2. Claude Opus 4.6 achieved an estimated task horizon of 718 minutes, successfully completing tasks requiring nearly a full human working day.
3. Month-long autonomous project completion by AI may become realistic around 2027, per current trend extrapolation from METR.
4. Visual computer-use tasks remain 40-100x behind software engineering tasks, indicating physical-world interaction as a persistent bottleneck.
5. Venture funds hold approximately $3 trillion in unrealized NAV, with 59% of unicorns now over 10 years old and 20% over 15 years old.
6. By end of 2025, DPI for venture funds in prime return years fell to approximately 12%, down from the historical ~20%.
7. Among 2021 funds, ~75% returned less than one-quarter of LP capital by year four, versus ~50% of 2017 funds at the same stage.
8. The 10 largest venture funds captured nearly 43% of all VC raised in 2025, and median US fund close time stretched to 15.3 months (from 9.7 months three years earlier).
9. AI accounted for over 50% of global venture deal value, with five companies (OpenAI, Anthropic, xAI, Scale AI, Project Prometheus) absorbing ~20% of total global VC funding.
10. The State of Web Dev AI 2026 survey of 7,200 developers found AI now generates 54% of code on average, up from 28% one year prior.
11. Claude Code received the highest positive sentiment score among coding agents, and Claude ranked first among AI tools that developers actually pay for.
12. Developers report spending between $50 and $500 per month on AI tools, with a majority agreeing the current AI market is a bubble.
13. Job displacement ranked as the most concerning AI risk among developers, ahead of military applications and misinformation.
14. Gokul Rajaram's shipping framework distinguishes four models based on product criticality and customer time availability: Square model (critical, time-strapped), Facebook model (non-critical, time-rich), Bundled releases, and Reliable and continuous.
15. Compliance in the US employs over 400,000 officers representing $40 billion annual labor costs, with 33,000+ new openings projected annually through 2036.
16. AI-native compliance opportunities include turning regulation into code (e.g., Tako for Brazil's labor regulations), replacing legacy systems (e.g., Sardine reducing Suspicious Activity Report preparation from 30+ minutes to under one minute), and augmenting workers with AI agents.

1. President Trump signed an executive order on Tuesday (presumably June 1 or 2, 2026) requiring certain AI companies to voluntarily submit new models for government testing 30 days before public release.
2. An earlier draft had mandated a 90-day voluntary review window, but industry insiders pushed for two weeks, leading to the 30-day compromise.
3. Trump delayed signing the more demanding version after pushback from venture capitalist and former White House AI czar David Sacks, citing competitiveness against China.
4. The order explicitly states it does not authorize mandatory governmental licensing, preclearance, or permitting for new AI models, including frontier models.
5. Trump signed the final version privately instead of with a planned attendance of top Silicon Valley CEOs.
6. The executive order directs the Department of Justice to prioritize enforcement against AI-assisted hacking and unauthorized access.
7. This order follows a previous Trump AI executive order from last December (2025) that mandated a single national AI policy framework to preempt state AI laws.