Jun 10, 2026 · AI & Tech Daily Digest
Today's AI / tech highlights for 2026-06-10, with summaries, links, and commentary.
I. Policy & Regulation
1. EU Emergency Order: Meta Must Restore Free WhatsApp Access for Rival AI Assistants by June 15 (Regulation / Antitrust)
Summary:
On Tuesday, June 10, 2026, the European Commission issued a rare interim measure requiring Meta to restore free WhatsApp access for AI chatbots from rivals such as OpenAI while an antitrust probe continues. Politico notes this is only the second time in 20+ years the EU has used such emergency power; the investigation began in December 2025 over whether Meta abused WhatsApp dominance to block third-party AI assistants. After effectively banning access in October 2025, Meta restored it for a fee in March 2026 — which the Commission says amounts to the same outcome as a ban. Competition Commissioner Teresa Ribera said the step prevents "serious and irreparable damage" to the general-purpose AI assistant market. Meta must comply by June 15 or face fines of up to roughly $20 billion (10% of annual revenue); the company plans to appeal, calling it "regulatory overreach" that subsidizes giant AI rivals at European businesses' expense.
Links:
- The Verge — WhatsApp ordered to host rival AI assistants for free
- The Straits Times — EU orders Meta to open WhatsApp to rival AI chatbots for free
Commentary:
As AI assistant competition shifts from model capability to messaging entry points, Brussels used a once-in-two-decades interim order to treat WhatsApp as critical infrastructure — Meta's walled-garden strategy hit an antitrust emergency brake harder than DMA alone.
2. China's MIIT Issues "AI + Information & Communications" Implementation Plan: 75% Urban 1ms Compute Coverage by 2028 (Policy / China)
Summary:
On June 10, 2026, China's Ministry of Industry and Information Technology (MIIT) published the Implementation Opinions on Innovative Development of "Artificial Intelligence + Information and Communications" (2026–2028). By 2028, the plan targets highly autonomous information and communications networks, 30+ high-value scenarios, flagship applications and specialized agents, and urban 1-millisecond compute latency circle coverage of at least 75%. By 2030, it envisions integrated sensing-computing-intelligence services and a mature collaborative innovation ecosystem. The document lays out 17 tasks covering AI-driven network architecture, large/small model coordination, multi-agent communication, compute backbone buildout, and a national integrated compute service system, plus dedicated action columns for network autonomy, intelligent compute networking, and intelligent compute services. Xinhua and Beijing Daily published same-day explainers framing the move as implementation of the State Council's "AI+" agenda.
Links:
- China Economic Net — Urban 1ms compute latency circle coverage expected to reach at least 75% within 3 years
- Beijing Daily — Seven Q&As + one chart: Understanding the "AI + Information & Communications" plan
Commentary:
Beijing upgraded the AI race from "build compute" to "hit the 1ms circle" — turning network latency and agent deployment into measurable national KPIs.
3. Apple Siri AI Blocked in the EU and Mainland China: DMA and Local Regulatory Hurdles (Product / Regulation)
Summary:
TechNode reported on June 10, 2026 that after Apple unveiled overhauled Siri AI at WWDC on June 8, the company confirmed the feature will not launch on iPhones or iPads in the EU or mainland China for now due to regulatory requirements. Siri AI supports multi-turn conversation using on-screen content, mail, messages, and photos, and can execute tasks across apps; developer betas shipped with iOS 27, with public betas expected later this year. Apple has criticized Digital Markets Act (DMA) interoperability demands; EU officials say they will not amend DMA for Apple. AI News noted on June 10 that the initial beta is English-only, China — Apple's most contested market — is fully excluded while domestic assistants face no such restriction, and "same-day global launch" for agents failed again at the regulatory border.
Links:
- TechNode — Apple delays AI-powered Siri in EU and China over regulatory hurdles
- AI News — Siri AI arrives with Google inside, and much of the world is locked out
Commentary:
Apple outsourced Siri's brain to Gemini but kept compliance in-house — WWDC's spotlight dimmed simultaneously in Brussels and Beijing, leaving blank spots on the agent distribution map from day one.
II. Models & Products
4. Decart Launches Oasis 3 Interactive World Model: Real-Time Photorealistic Driving Environments via API at $0.02/second (Models / Physical AI)
Summary:
TechCrunch reported exclusively on June 10, 2026 that two-year-old startup Decart released Oasis 3, an interactive world model generating physically accurate multi-camera (front + side) photorealistic driving environments in real time via API, initially targeting AV firms simulating rare scenarios and planning expansion into robotics and other physical AI. Decart aims to build a developer ecosystem around world models akin to OpenAI's language-model playbook; API pricing is $0.02 per second, with enterprise rates by use case. The company cites a 100,000+ developer community built on its real-time video model Lucy in e-commerce and live streaming. Oasis 3 arrives weeks after Decart raised $300 million at nearly a $4 billion valuation, with strategic investors including Toyota, Adobe, eBay, and NVIDIA.
Links:
Commentary:
World models moved from paper demos to per-second billing — Decart's bet is that physical AI's "OpenAI moment" depends on developers generating infinite edge cases, not one impressive video clip.
5. Waymo Open-Sources ReD (Reference Driver) Cognitive Model: A "Behavioral Crash Test Dummy" for AV Safety Benchmarks (Autonomous Driving / Open Source)
Summary:
The Verge reported on June 10, 2026 that Waymo published a Nature Communications paper and open-sourced ReD (Reference Driver), a cognitive model explaining how human drivers make split-second crash-avoidance decisions. Built with Delft University of Technology, ReD works like a behavioral crash-test dummy — evaluating whether AV systems avoid danger rather than only assessing hardware after impact. Waymo says it is collaborating with researchers, regulators, and standards bodies including SAE to establish consensus on a "careful and competent" human response baseline, and has made ReD publicly available for testing.
Links:
Commentary:
When regulators ask whether robotaxis beat human drivers, Waymo first open-sourced a reproducible human-behavior benchmark — safety narratives shift from mileage counts to auditable behavioral comparisons.
6. Zilliz Launches Vector Lakebase Public Preview: Vector Database Extends into Unified AI Data Platform (Infrastructure / Product)
Summary:
PR Newswire reported on June 10, 2026 that Zilliz — behind open-source Milvus — announced public preview of Zilliz Cloud Vector Lakebase. The release keeps real-time vector search at the core while adding shared lake-native storage and on-demand compute, unifying real-time serving, interactive discovery, and large-scale batch analytics on one data foundation where all workloads share a single logical data copy and compute is billed only when active. Zilliz says Zillow, OpenEvidence, Exa, MiniMax, and 10,000+ enterprises and AI teams already rely on its vector engine; preview spans 30+ regions across AWS, GCP, and Azure.
Links:
Commentary:
The next RAG stack battle isn't bigger embedding models — it's merging vector search, data lakes, and analytics into one billing curve; Zilliz answers "who defines the AI data layer" with Lakebase.
7. Moore Threads Open-Sources MusaCoder: First Full-Stack Code Model Trained on Domestic GPU Infrastructure (Open Source / China)
Summary:
Sina Tech reported on June 10, 2026 that Moore Threads released and open-sourced MusaCoder, a code model specialized for GPU low-level operator generation in 9B and 27B sizes, supporting automatic generation of high-performance CUDA/MUSA native kernels from standard PyTorch operators. Moore Threads says it is the first open code model fully trained and validated on domestic GPU infrastructure, with SFT, RFT, RL, and online compile-verify pipelines completed on MTT S5000 Kua'e compute clusters. On KernelBench, MusaCoder-27B-RL scored Overall Pass@8 93.2% and Avg.@8 88.60%, which the company says exceeds mainstream models including Claude Opus 4.7 and DeepSeek-V4 Pro.
Links:
Commentary:
China's next AI hurdle isn't "can we train big models" but "can we train kernel-writing models on our own GPUs" — MusaCoder pushes compute sovereignty down to the compiler stack.
8. Tsinghua Team's DrugCLIP Published in Science: AI Drug Virtual Screening Up to 10 Million× Faster (Research / China)
Summary:
36Kr reported on June 10, 2026 that a Tsinghua-led team proposed DrugCLIP, an "AI contrastive deep learning" framework for ultra-fast, high-precision virtual screening — up to 10 million× faster than traditional molecular docking in reported scenarios (with 100,000× cited in some benchmarks) — published in Science. The team built GenomeScreenDB covering nearly 10,000 human targets, 20,000+ binding pockets, and 2 million+ potential hit molecules with open structure, docking pose, and scoring data. Wet-lab results showed 15% hit rate against the norepinephrine transporter and 17.5% against TRIP12, a target lacking full ligand structures, using only AlphaFold2 predicted structures.
Links:
Commentary:
Post-AlphaFold drug discovery competition shifts from "predict structures" to "screen the whole genome fast enough" — DrugCLIP converts compute advantage directly into a druggable-genome map.
III. Infrastructure & Capital
9. Meta Signs First India AI Data Center Deal with Reliance: 168MW Jamnagar Project (Infrastructure / India)
Summary:
TechCrunch reported on June 10, 2026 that Meta and Indian conglomerate Reliance Industries will build a 168-megawatt AI-enabled data center in Jamnagar, Gujarat, expected ready within two years and expandable over time. It is Meta's first AI infrastructure bet in India, extending prior multi-billion-dollar investment in Jio Platforms and a $100 million enterprise AI JV launched in 2025. Reliance says the facility will plug into Meta's global AI compute network; Meta separately contracted nearly 1 gigawatt of new renewable capacity in India via CleanMax and Fourth Partner Energy. Microsoft, Amazon, Google, OpenAI, and Uber have also announced recent India AI/cloud infrastructure investments.
Links:
Commentary:
As U.S. giants face data-center moratoriums at home, India becomes the default expansion valve — Meta uses Reliance to splice Jamnagar into the global AI backbone.
10. SpaceX IPO Demand Exceeds $250 Billion: Roughly 3.5–4× the $75 Billion Offering Size (Capital / IPO)
Summary:
Reuters, via CNBC TV18 and Technology.org on June 10, 2026, reported that investor demand for Elon Musk's SpaceX IPO has surpassed $250 billion — about 3.5 to 4× the $75 billion the company seeks to raise. Long-only funds reportedly placed sizable orders; Musk briefly joined some Zoom roadshow sessions. SpaceX has publicly set a fixed $135/share price implying roughly a $1.77 trillion valuation; pricing is expected Thursday, June 11, with SPCX trading on Nasdaq starting June 12. Reports also cite SpaceX pitching a $23 trillion space-based AI compute opportunity; subscription figures reflect indications of interest, with final allocations set at pricing.
Links:
- CNBC TV18 — SpaceX IPO demand is approaching four times oversubscribed
- Technology.org — SpaceX IPO Demand Hits $250B, 4x Oversubscribed
Commentary:
Four-times oversubscription doesn't guarantee four-times day-one gains, but it proves that in a year when Anthropic and OpenAI queue for IPOs, capital still pays a premium for the rocket + Starlink + space-compute narrative.
11. China Plans ~CNY 2 Trillion National AI Data Center Network Over 5 Years (Policy / Infrastructure / China)
Summary:
Lianhe Zaobao reported on June 10, 2026, citing Bloomberg, that China's NDRC and other agencies are drafting a blueprint to invest roughly CNY 2 trillion (~$295 billion) over five years in a nationally interconnected AI data center network operated mainly by China Mobile, China Telecom, and other SOEs, requiring at least 80% of key hardware and software (including AI chips) from domestic suppliers such as Huawei. Sources say the plan remains preliminary; including new power grid integration, total investment could exceed CNY 5 trillion, targeting 2028 integration of dispersed facilities into a unified compute network to accelerate AI in healthcare, transport, and urban management. The initiative is part of March's "Six Networks" plan (~CNY 7 trillion total).
Links:
- Lianhe Zaobao — China to invest CNY 2 trillion nationwide on data centers over five years
- Sina Finance — CLSA: China launches CNY 2 trillion national AI infrastructure blueprint
Commentary:
CNY 2 trillion buys more than racks — it's an "80% domestic" compute network, escalating the U.S.–China AI infrastructure race from chip export controls to national network architecture.
IV. Ecosystem & Commerce
12. WeChat AI Ecosystem Onboarding Wave: Meituan, JD.com, Trip.com Among First Beta Partners; Tencent Shares Up Nearly 5% (Ecosystem / China)
Summary:
Cailian Press reported on June 10, 2026 that WeChat published developer guidance on June 8 via its official account, offering "auto mode" (platform reads source code; AI operates pages directly) and "dev mode" (businesses wrap SKILLs for WeChat AI to invoke), currently in closed beta. Meituan, JD.com, Trip.com, Tongcheng, Poizon (Dewu), and Midea announced first-batch beta integration covering food delivery, e-commerce, travel, and smart-home control; Alibaba's Qwen also opened third-party Agent access on June 9. Tencent Holdings shares rose nearly 5% in early trading on June 10. Guidance stresses integration is voluntary and won't affect existing mini-program services; dev-mode code is not yet eligible for formal app review.
Links:
- Caijing China — WeChat AI ecosystem opens to developers; leading firms join beta
- WeChat Open Platform — Mini Program AI Dev Mode (beta) Integration Guide
Commentary:
Super-app AI wars shifted from "who has the model" to "who can wrap takeout, flights, and appliances as SKILLs" — WeChat and Qwen opened Agent interfaces the same week, repackaging China's consumer internet as task-completion infrastructure.
Today's Summary
- EU AI entry-point battle escalates: On June 10, the Commission ordered Meta to restore free WhatsApp access for rival AI assistants by June 15, a rare interim measure with fines up to ~$20 billion.
- U.S.–China compute policy moves in parallel: Beijing published the "AI + Information & Communications" plan (75% 1ms urban coverage by 2028) and is reportedly advancing a CNY 2 trillion national data center blueprint; Meta signed a 168MW AI data center deal in India.
- Physical AI and AV toolchain: Decart Oasis 3 world-model API launched; Waymo ReD open-sourced as a behavioral benchmark — simulation and evaluation infrastructure strengthened the same day.
- SpaceX IPO enters pricing countdown: June 10 demand exceeded $250 billion, roughly 4× oversubscribed, with June 11–12 pricing and listing ahead.
- China industry side: Moore Threads MusaCoder open-sourced, DrugCLIP in Science, and WeChat AI ecosystem beta onboarding by major merchants — models, compute, and super-apps advancing together.
Daily Framing:
June 10, 2026 reads as an "AI entry points vs. infrastructure collision day" — Brussels fought for assistant distribution on WhatsApp, Beijing and Washington's supply chains drew competing compute maps, and SpaceX ran the final roadshow for a space-plus-AI capital narrative.
This digest is compiled from real-time search results and is for reference only; facts are subject to original sources.
Date: June 10, 2026 (Wednesday)