SpaceX Acquires Cursor for $60 Billion in All-Stock Deal
In the week's biggest AI story, SpaceX filed merger documents on June 17 to acquire Cursor, the AI-powered coding assistant, in an all-stock deal valued at roughly $60 billion. The announcement sent SpaceX shares soaring 17%, briefly making it the fourth most-valuable US company.
The deal caps a meteoric rise for Cursor, which went from an OpenAI accelerator project to a $4 billion annual recurring revenue business in just two years. But the acquisition also comes at a moment of vulnerability — Cursor's market share had slipped from 41% to 26% over the past twelve months as deep-pocketed rivals closed in.
Perhaps most intriguing: the two companies have been jointly training a coding model on xAI's Colossus supercomputer, expected to launch across both platforms soon. With this deal, every major AI coding assistant is now owned by a tech giant — effectively closing the door on the independent startup path for similar tools.
The acquisition also has direct competitive implications for Anthropic's Claude Code, which now faces a far better-resourced rival backed by SpaceX's capital and compute infrastructure.
Fable 5 Enters Second Week Offline Under US Export Controls
Anthropic's frontier model Claude Fable 5 — along with Claude Mythos 5 — remains completely offline as of June 19, marking a full week since the US Commerce Department issued an extraordinary export control directive on June 12.
The directive, which Anthropic called "legally binding," requires the suspension of all access to both models for any foreign national, whether inside or outside the United States — including Anthropic's own foreign-national employees. The shutdown was triggered at 5:21 PM ET on June 12, after reports surfaced that another company had successfully jailbroken Mythos's cybersecurity safeguards, alarming administration officials about potential national security risks.
The situation grew more complex at the G7 summit, where diplomatic pressure has reportedly complicated restoration negotiations. Meanwhile, the US Court of Appeals for the DC Circuit denied Anthropic's request for an emergency stay of the Pentagon's separate supply chain risk designation.
Anthropic's other models, including the recently crowned Claude Opus 4.8 — which holds the #1 spot on the Artificial Analysis Intelligence Index with a score of 61.4 — remain available.
G7 Leaders Meet AI Executives for Safety Talks
The G7 AI Working Lunch concluded on June 17, bringing together leaders from OpenAI, Anthropic, and Google DeepMind with G7 heads of state to discuss "safe, rapid, and effective AI deployment" and youth protection online.
The outcome: a set of non-binding voluntary safety commitments, with youth safety protection as the primary focus. The framework reflects the US administration's preference for voluntary standards over mandatory regulation — a position that stands in contrast to the EU's risk-based AI Act approach.
Separately, France's President Macron secured a €45 billion commitment from SoftBank for AI data center infrastructure, positioning France as Europe's leading AI hub. Mistral AI's presence at the discussions underscored Europe's push for alternatives to US-dominated AI development.
OpenAI Says GPT-5.5 Now Outperforms Physicians
OpenAI announced on June 18 that GPT-5.5 Instant now outperforms physician-written health answers across multiple evaluation criteria. A panel of physicians reviewed 3,500 model outputs against responses written by doctors with unlimited time and internet access.
The physicians rated GPT-5.5 Instant higher than both older models and human-written answers on accuracy, communication, completeness, and health decision helpfulness, according to OpenAI. The model also shows improvements in recognizing when urgent care may be needed and explaining uncertainty without overstating confidence.
With more than 230 million people already using ChatGPT for health and wellness questions, and GPT-5.5 Instant available to all free-tier users, the stakes of these claims are enormous. The healthcare AI space remains a contentious frontier — impressive benchmarks notwithstanding, real-world clinical validation still lags behind.
All Eyes on Gemini 3.5 Pro as Prediction Markets Bet on Late June Launch
Google's Gemini 3.5 Pro remains the most anticipated model release of the month. Announced at Google I/O on May 19 by Sundar Pichai, who told the audience to "give us until next month," the model has yet to go GA — and is still limited to Vertex AI enterprise preview.
Polymarket prediction markets are concentrating odds on June 23 and June 30 as the most likely launch windows, reflecting typical end-of-quarter timing. The model is expected to feature a 2 million token context window and Deep Think reasoning, targeting the use cases that Gemini Ultra once covered.
Meanwhile, Gemini 3.5 Flash — which shipped at I/O and is already the default in the Gemini app and AI Mode in Search — continues to prove that last generation's Pro-level performance is this generation's baseline. Google also announced the Antigravity CLI on June 18 as the replacement for the Gemini CLI, requiring developers to migrate their workflows.