Anthropic model shutdown sparks backlash & G7 debates trusted AI access - Tech News (Jun 18, 2026)

[0:33] Anthropic model shutdown sparks backlash

We’ll start with the AI policy shockwave. Anthropic employees are reportedly rattled after the U.S. government abruptly ordered the company to pull access to its latest models, Fable 5 and Mythos 5, citing national security. The scramble wasn’t just about speed—reports describe shifting explanations, from concerns about foreign access to claims of a security weakness, with executives pushing for clarity and getting little in return. What makes this especially notable is that similar capabilities exist elsewhere in the market, which is fueling a growing argument that enforcement is turning unpredictable. For businesses building on frontier models, that kind of uncertainty can be as disruptive as any technical limitation.

In health tech, Paradromics and University of Michigan Health reported the first human implantation of the company’s Connexus brain-computer interface as part of an FDA-approved feasibility study. The goal is to evaluate long-term safety and whether it can help restore speech or enable computer control for people with severe motor impairments. This is meaningful because it moves high-bandwidth, implanted BCIs further into real clinical timelines—where durability, follow-up, and day-to-day usability matter more than lab demos.

Autonomy is also getting more concrete on the streets. Mobileye says it plans to launch its own standalone robotaxi service in the U.S. in 2027. The shift here is strategic: Mobileye has long been a supplier of driver-assistance and autonomy tech, but running a consumer-facing service puts it in direct competition with operators already trying to prove robotaxis can work at scale. It also suggests a broader trend: autonomy vendors increasingly want end-to-end businesses, not just licensing revenue.

And autonomy isn’t only about vehicles. Nvidia researchers, along with university partners, described a framework that lets teams of AI coding agents run robotics training experiments end-to-end—planning runs, debugging failures, and iterating with less human involvement. The headline isn’t that robots suddenly became magical; it’s that the bottleneck is shifting. If experiments can be launched and refined faster, progress speeds up—but costs, coordination, and wasted time when robots sit idle still matter. It’s the early shape of what a “self-running” robotics lab might look like.

Finally, two space stories with very different flavors. NASA selected Relativity Space—now led by Eric Schmidt—to build and launch a Mars mission called Aeolus, meant to deliver daily global data on Mars’ atmosphere. It’s an ambitious public-private bet with a tight schedule, and it puts pressure on Relativity to execute with hardware that still has a lot to prove. At the same time, markets are buzzing with speculation about a possible mega-merger between SpaceX and Tesla after SpaceX’s blockbuster IPO. It’s far from a done deal, but the governance questions are immediate: Musk would, in effect, be negotiating with himself, and regulators would inevitably scrutinize a combined entity spanning space, communications, AI, and robotics. Even if it never happens, the chatter reflects how investors are thinking—less in terms of single companies, and more in terms of vertically integrated tech empires.