The Signal — June 13, 2026
The US government has inserted itself directly into AI access policy, and the ripple effects are already reshaping who builds with what, and where.
US Government Orders Anthropic to Suspend Foreign Access to Fable 5 and Mythos 5
Two days after we covered Anthropic's voluntary walkback of Fable 5 throttling, the situation has escalated sharply: the US government has ordered Anthropic to suspend access to both Fable 5 and Mythos 5 for foreign nationals, citing national security concerns. What started as a company product decision is now a government-mandated restriction on frontier AI access.
Anthropic confirmed the directive but pushed back on the framing, calling the order a "misunderstanding" and saying it is working to restore access. The move opens a new front in US AI policy, where export-control thinking is being applied not just to chips but to model access itself. Reuters reports the restriction applies broadly, while TechCrunch notes the irony: Anthropic's own safety communications may have drawn the government's attention to its models' capabilities. Le Monde frames this as a Trump administration national security action with global implications for AI availability.
Sources: Anthropic · Reuters · TechCrunch · Le Monde
SpaceX Rents Colossus 1 Data Center to Anthropic After xAI Grok Struggles
In one of the more unexpected infrastructure moves of the year, SpaceX is renting its Colossus 1 data center to Anthropic after xAI's own teams struggled to effectively use it for Grok model development. The facility, originally built to power Elon Musk's AI ambitions, is now serving a direct competitor.
According to Bloomberg, xAI's internal teams had difficulty utilizing Colossus 1 at scale, making it more economical to lease the capacity than let it sit idle. Anthropic (now simultaneously fighting a government access order and scaling its compute) gets a capacity injection at a useful moment. The deal makes a point that keeps recurring in AI infrastructure: building the data center is the easy part. Operating it at frontier scale is where the real difficulty lives.
Sources: Bloomberg · Yahoo Finance
Meta's AI Unit Faces Internal Turmoil as Engineers Revolt
Meta's AI reorganization is hitting a wall from the inside. Engineers in the company's months-old AI unit are describing the experience as "soul-crushing," according to TechCrunch, with reports of low morale and organizational dysfunction. Separately, Wired reports that employees are actively resisting Zuckerberg's mandate for a companywide AI hackathon.
Together, the two reports suggest Meta's AI push is generating real cultural friction. Reorganizations are never smooth, but when your own engineers reach for the word "gulag," that goes beyond normal transition pain. For a company betting its future on AI, talent retention becomes a concrete problem, especially when Google, OpenAI, Anthropic, and a dozen well-funded startups are all hiring hard.
Sources: TechCrunch · Wired
On the Editor's Desk
We held several stories today. The NVIDIA Blackwell AgentPerf benchmark and TCS-Anthropic partnership both scored well on significance but were already covered in previous editions. A story on Anthropic blindsiding business partners was also held as prior coverage. An Axios/Reuters piece on Anthropic's "misunderstanding" framing was folded into our lead rather than run separately. Today's three picks reflect genuinely new developments: a government action with no precedent in AI policy, a surprising infrastructure deal that reveals cracks in xAI's execution, and mounting evidence that Meta's AI strategy is alienating its own workforce.
