Anthropic’s Fable 5 and Mythos 5 Go Dark Under Export Order The Trump administration ordered Anthropic to block access to its two most powerful models for all foreign nationals — including users inside the US and the company’s own foreign-national employees — after Amazon researchers discovered a jailbreak. Anthropic shut the models down and is negotiating to restore access, while G7 leaders cited the blackout as proof that American AI kill-switch authority is a geopolitical liability.
🏛️ Policy & Geopolitics
Anthropic got hit by export rules nobody understands — The Verge
Commerce Secretary Howard Lutnick directed Anthropic to apply export controls to Fable 5 and Mythos 5, treating access by any foreign national — even inside the US — as a controlled export. The trigger was a jailbreak of Mythos discovered by Amazon researchers, which alarmed national security officials. Anthropic shut the models down entirely and says it believes the action is a misunderstanding; talks to restore access are ongoing.
World leaders want American AI. They just don’t want America to be able to turn it off. — TechCrunch
French President Macron and Indian PM Modi raised formal alarms at the G7 summit about US sovereign control over frontier AI access, warning that any country relying on American AI models is one executive order away from losing that capability. The Anthropic blackout — which affected customers globally mid-week — turned a theoretical concern into a live case study.
Vibe-decoding the White House-Anthropic fight over Fable — The Verge: Analysis of the political dynamics behind the export dispute, including the administration’s broader push to assert control over frontier AI before it can be weaponized.
Anthropic, Trump Officials Seek Deal on Restoring Powerful Model Access — WSJ: Both sides are working quickly toward a resolution; the White House faces pressure to show it can oversee the industry responsibly without crippling it.
Two-thirds of Americans think AI is advancing too quickly — The Verge: Pew Research finds 49% of Americans now use AI chatbots, up from 33% in 2024, yet 63% say the technology is advancing too fast and only 16% expect a positive societal impact.
🚀 Industry Moves
SpaceX to acquire the AI coding startup Cursor for $60 billion — CNBC
SpaceX — which absorbed xAI and the X platform into a division called SpaceXAI in February — is buying Anysphere (Cursor) for $60 billion in stock, the largest AI startup acquisition on record. Cursor has scaled to roughly $2.6 billion in annualized revenue, more than one million paying users, and claims deployment across 64% of Fortune 500 companies. The deal is expected to close in Q3 pending regulatory approval.
Leaked financial docs show OpenAI is losing billions of dollars a year — Ars Technica
Audited financials obtained by an independent journalist show OpenAI’s revenue surged from $3.7B (2024) to $13.07B (2025) — but expenses grew even faster, from $7.81B to $19.18B, widening net losses despite record revenue. The company has told investors it expects to reach profitability by 2030.
Anthropic “pauses” token-based billing for its Claude Agent SDK — Ars Technica: Anthropic reversed pricing changes that would have billed Agent SDK usage separately, just before they were set to take effect. Outside SDK usage reverts to standard API rates while Anthropic redesigns its plan structure.
Anthropic becomes first AI startup to join the Frontier carbon removal coalition — TechCrunch: Anthropic joined the Frontier coalition alongside $915M in new advance purchase pledges for carbon removal projects.
Anthropic’s latest feud with the Trump admin may actually help it, sales data suggests — TechCrunch: Ramp spending data shows Anthropic’s enterprise adoption growing strongly; analysts suggest the government standoff may be signaling Fable/Mythos’ capability to enterprise buyers.
🛠️ Developer Tools & Models
Cursor Origin — Cursor (via X)
Hours before the acquisition news broke, Cursor announced Origin — a Git-compatible forge designed from the ground up for parallel AI agent workflows. Where GitHub was built for human-scale development, Origin is architected for dozens of agents branching, committing, reviewing, and rebasing simultaneously.
Android 17 Expands AI Agent Integration — Android Developers Blog
Android 17 ships AppFunctions and Android MCP, enabling apps to expose tools that on-device agents can discover and execute. Google is positioning this as the foundation of an “intelligence system” — a platform shift from apps that respond to taps, to apps that agents can orchestrate on the user’s behalf.
Android 17 starts hitting Pixel phones and watches today — Ars Technica: The rollout is live for Pixel devices; the release adds a modest feature set while laying infrastructure groundwork for the AI agent layer.
GLM-5.2 — Z.ai: Z.ai launched GLM-5.2 with a 1 million-token context window and coding-first design for agentic software engineering tasks; MIT-licensed open weights are planned for release next week.
OpenAI released CDP support for browser use on Codex — TestingCatalog: Codex can now access live browsers via Chrome DevTools Protocol, enabling real-time JS profiling and DOM modification; the feature is opt-in and excludes EEA, UK, and Switzerland.
OpenAI prepares major ChatGPT voice upgrade with GPT-Bidi-1 — TestingCatalog: GPT-Bidi-1 is a bidirectional audio model for voice mode that can listen and speak simultaneously, absorbing interruptions and adjusting mid-sentence.
Why Weibo’s tiny VibeThinker-3B has the AI world arguing over benchmarks again — VentureBeat: The 3B-parameter model posts coding scores in the same range as Claude Opus 4.5, reviving the benchmark contamination debate.
📱 Hardware & Physical AI
Snap unveils $2,195 AR glasses as CEO Evan Spiegel bets on post-smartphone future — CNBC
Snap’s Specs AR glasses — 12 years in development — will ship later this year in the US, UK, and France at $2,195 with a $200 refundable preorder deposit. They overlay digital visuals onto the user’s field of view for shared real-world computing experiences, entering a market where Meta’s Ray-Ban line has found early traction and Google is preparing a competitor.
Qualcomm wants to be the chip inside whatever replaces your smartphone — TechCrunch: Qualcomm announced Snapdragon Reality Elite for mixed-reality glasses and the STAR (Scalable Turnkey AI-Ready) toolkit for AI wearables, targeting 40+ devices in development.
The next humanoid robot might not look human at all — The Verge: Genesis AI (backed by ex-Google CEO Eric Schmidt) unveiled Eno — a wheeled, foldable robot that can reason and adapt beyond predefined tasks — alongside an LG partnership for industrial deployments by year-end.
Google’s first smart speaker in six years arrives next week — TechCrunch: The $99 Google Home Speaker with Gemini ships June 25, replacing rigid Assistant commands with conversational AI interactions; preorders open today.
NVIDIA Blackwell Sweeps MLPerf Training 6.0 — NVIDIA Blog: Blackwell GPUs dominated MLPerf Training 6.0, setting records at 8,192-GPU scale using NVLink and NVFP4 quantization; reliability tools including the RAS Engine ensure fast recovery in large clusters.
Apple Plans Camera AirPods Alongside Upgraded Foldable iPhone in 2027 — Bloomberg: Three products — camera AirPods, a second-generation foldable iPhone, and a 20th-anniversary iPhone — are in advanced development for a coordinated late 2027 launch Apple is positioning as its biggest-ever product wave.
Generated by claude-sonnet-4-6 on 2026-06-17T10:00:00Z