It is Thursday, June 19, 2026. Fable 5 is still offline but the people who shut it down and the people who built it are finally in the same room. SpaceX spent $60 billion on a code editor four days after the largest IPO in history. And the reason Anthropic’s two best models are dark turns out to involve a South Korean phone company, an Amazon vulnerability report, and Dario Amodei refusing a government ultimatum. Here is everything.
The real story behind the Fable 5 ban: a Korean telecom company, an Amazon report, and Dario said no to both options
WIRED and The Washington Post have now reported the full sequence of events that produced the June 12 export control ban, and the story is considerably more specific than anything the government or Anthropic disclosed at the time.
It happened in two steps. Step one: the White House identified SK Telecom — South Korea’s largest wireless carrier, a $100 million Anthropic investor since 2023, and a Project Glasswing partner with access to Claude Mythos 5 — as a company suspected of having ties to China. The administration asked Anthropic to revoke only SK Telecom’s access. Anthropic complied the same day. Step two, the same week: Amazon researchers separately identified potential vulnerabilities in Fable 5 and reported them to the White House. The administration, already suspicious about Anthropic’s ability to control access, concluded it could not trust Anthropic to safeguard its most advanced AI technology. The export control letter ordering all foreign national access revoked arrived at 5:21 pm on June 12.
The David Sacks layer, disclosed on June 13: before issuing the ban, the White House gave Dario Amodei a choice. Fix the jailbreak or voluntarily de-deploy the model. Dario refused both options. Anthropic’s public position is that the identified jailbreak — asking the model to analyze code for security vulnerabilities — is a standard capability already present in GPT-5.5 and used daily by security teams. The administration issued the directive anyway.
SK Telecom has stated clearly it has no ties to China and does not use Huawei or ZTE equipment in its core networks. The White House has not published the specific evidence behind its determination. The Korean government is now formally involved, which is why Anthropic chose this week — the week of the ban — to open its Seoul office.
Read the full reporting: LLMBase analysis of the WIRED reporting on the SK Telecom trigger | Korea JoongAng Daily on the White House pinning the block on Korean telecom | LLMBase on the White House demand to eliminate all jailbreaks before relaunch | ExplainX on the Sacks ultimatum and restoration paths
Anthropic opened its Seoul office this week and its executive said Fable 5 returns “within days”
Anthropic formally opened its Seoul office on June 17 and 18 — its third in Asia-Pacific after Tokyo and Bengaluru — at the Conrad Hotel in Yeouido. The event was intended to celebrate Korean enterprise adoption. It was dominated by questions about the export control ban on the models that powered much of that adoption.
The most significant statement came from Chris Ciauri, Anthropic’s Managing Director of International: “We are very confident that in the coming days, the models will become available again.” That is the most specific positive signal Anthropic has given since the June 12 shutdown. Previous communications used open-ended language. Plural “days” with expressed confidence suggests active negotiations are producing movement.
The commercial announcements at the Seoul opening were substantial regardless of the ban. NAVER — Korea’s largest web portal — deployed Claude Code across its entire engineering organization. Samsung SDS is deploying Claude Cowork and Claude Code across Samsung Electronics. LG CNS is rolling Claude across LG Group. Nexon is using Claude Code for live-service game development. Hanwha Solutions is deploying Claude globally through AWS Bedrock. Channel Corp is powering Channel Talk — used by 230,000 businesses — on Claude.
Anthropic also signed a Memorandum of Understanding with Korea’s Ministry of Science and ICT committing both parties to AI safety collaboration, Korean-language model evaluation with the Korea AI Safety Institute, and government-to-government information sharing on AI-enabled cyber threats. The MOU creates a formal diplomatic channel that the White House situation will inevitably travel through.
Read the coverage: Anthropic’s official Seoul office announcement | Korea JoongAng Daily on Ciauri’s “within days” statement | UPI on the Seoul opening and its diplomatic context
SpaceX spent $60 billion on a code editor four days after the largest IPO in history
On June 16 — four days after trading opened — SpaceX announced it is acquiring Anysphere, the San Francisco startup behind the AI coding assistant Cursor, in an all-stock transaction valued at $60 billion. It is the largest acquisition of a venture-backed startup on record. SpaceX shares rose 8 percent on the announcement, pushing the company’s market cap past $2.7 trillion — ahead of both Amazon and Meta.
Cursor was founded in 2022 by four MIT classmates and had raised $3.4 billion from Andreessen Horowitz, Thrive Capital, Accel, and Coatue. It crossed $1 billion in annualized revenue in November 2025. The $60 billion price is roughly a 2x premium on Cursor’s last private valuation and represents a 17x return on total invested capital. The deal includes a $10 billion break-up fee.
The strategic logic: SpaceX identified a $26 trillion addressable market in AI in its IPO documentation, and Cursor is the most-used AI coding tool among professional developers. The acquisition gives SpaceX’s xAI division — which has Grok and the Colossus compute cluster but limited product distribution — a high-frequency developer workflow used daily by millions of engineers. SpaceX and Cursor disclosed they have already been jointly training a new AI model using Colossus 2’s compute. Cursor CEO Michael Truell said the goal is to build the world’s most useful AI models. Joint AI model training is already underway. Closing is expected in Q3 2026.
The competitive read for Claude Code and OpenAI Codex is direct. SpaceX now has compute infrastructure through Colossus, a frontier AI model through xAI, and daily developer distribution through Cursor. That is the same stack Anthropic is building with Claude Code and the same stack OpenAI is building with Codex. It is now a three-way race with a rocket company in it.
Read the coverage: CNBC on the deal announcement and SpaceX’s market cap crossing Amazon and Meta | CBS News on the terms and Cursor CEO Michael Truell’s statement | Crunchbase on why this is the largest venture-backed startup acquisition on record | MLQ on what the deal means for the AI coding market
OpenAI acquired the tools that every Python developer uses and the open-source community is nervous
OpenAI is acquiring Astral — the company behind uv and ruff, two of the most dominant Python developer tools of the last three years. uv is the fastest Python package installer available and has broadly displaced pip and pip-tools. ruff has displaced flake8 and pylint as the default Python linter across major open-source projects. Together they are embedded in the daily workflow of millions of Python developers worldwide.
The stated goal is integrating both tools into Codex, OpenAI’s AI coding platform. The strategic logic is the same as the SpaceX-Cursor deal: control a critical link in the developer workflow. If uv and ruff are the default Python toolchain inside Codex-managed environments, OpenAI is embedded at the infrastructure level of every Python project Codex touches.
The community concern is about what happens to the open-source licensing and governance of both tools now that they belong to OpenAI. Astral built its reputation specifically on being fast, independent, and developer-first. The acquisition terms, roadmap, and licensing commitments have not been fully disclosed. The developer community response on Hacker News and GitHub Discussions has been cautious. What OpenAI decides to do with the open-source governance of uv and ruff will be closely watched.
Read the coverage: The New Stack on the Astral acquisition and what it means for Codex
Also worth reading today
- Gemini 3.5 Pro is still not out. Eleven days remain in June. Sundar Pichai promised it last month on stage at Google I/O. Polymarket traders are currently pricing roughly 50-55% odds of a release before June 30. (Polymarket live odds)
- Someone built a website that pings Anthropic’s API every minute to check whether Fable 5 is back. The answer as of this morning is still no. (IsFable5Back.com)
- June 20 is the refund processing cutoff for customers who paid for Fable 5 usage credits now sitting offline. If you are affected, contact Anthropic support before tomorrow. (ExplainX Fable 5 status tracker)
- Google released its first smart speaker in six years, featuring Gemini AI as the built-in assistant. The previous Google smart speaker was the Nest Audio in 2020. All three major smart home platforms — Google, Amazon, and Apple — are now racing to upgrade their voice AI simultaneously. (LLMBase news tracker)
- MiniMax M3 open weights are drawing serious enterprise evaluation in the wake of the Fable 5 ban, specifically because open-weight models cannot be recalled by any government directive. The ban converted an abstract data sovereignty argument into a concrete production risk event overnight. (VentureBeat on enterprise alternatives and the open-weights advantage)
That is your Thursday. Fable 5 is coming back in days, not weeks, according to the person Anthropic sent to Korea. SpaceX bought the code editor everyone uses four days after going public. OpenAI bought the Python tools everyone uses and has not explained what happens to the open-source licenses. And Gemini 3.5 Pro has eleven days left to make Sundar Pichai’s promise true. See you tomorrow.