The week of June 15 was shaped by the collision of frontier AI with geopolitical power, alongside a pair of deals that suggested the industry's consolidation is accelerating.
The dominant story ran across three interconnected events. The Trump administration issued a directive suspending foreign access to Anthropic's Fable 5 and Mythos 5 models on national security grounds — citing an alleged jailbreak that Anthropic publicly contested. A CNBC analysis framed the episode as a cautionary tale: Anthropic had called for AI regulation and received a unilateral executive action with no statutory process and no prior consultation. By midweek, the same cast of characters — Anthropic CEO Dario Amodei, OpenAI's Sam Altman, and Google DeepMind's Demis Hassabis — attended the G7 in France as official participants rather than observers, a first for frontier AI leaders. Amodei and Hassabis jointly called for a U.S.-led coalition on global AI safety standards; the Mythos dispute suggested Washington had already decided to act first and consult later.
SpaceX announced it will acquire Anysphere, the maker of Cursor, for $60 billion in stock — announced days after SpaceX's blockbuster IPO and following its earlier merger with xAI. The deal positions SpaceX with a vertically integrated AI engineering stack. Cursor simultaneously announced Origin, an agent-first git hosting platform designed from the ground up for machines as primary collaborators rather than humans, with a launch expected in fall 2026.
OpenAI shipped GPT-5.5 on Thursday, continuing its rapid GPT-5 family cadence and making the model available across the API, Codex, and ChatGPT. The company also announced that its models, Codex, and Managed Agents are now available natively on AWS, broadening enterprise distribution beyond Azure. OpenAI also introduced LifeSciBench, an expert-authored benchmark for life science research tasks developed alongside practicing scientists.
On the secondaries: Salesforce acquired Fin — the AI customer service agent formerly known as Intercom — for $3.6 billion to strengthen Agentforce. Microsoft's Work IQ APIs reached general availability, giving enterprise agents programmatic access to a workplace intelligence layer built from email, calendar, meetings, and people signals. Anthropic opened a Seoul office with enterprise deployments at NAVER, Samsung SDS, and LG CNS. Google launched Gemini CLI GitHub Actions, an open-source agent that can fix bugs and implement features when mentioned in a pull request. To watch next week: the outcome of Anthropic's scheduled meeting with White House officials over the Mythos directive, and whether the administration's approach becomes a template for other labs.