TL;DR: Despite an ongoing legal battle with the Pentagon, Anthropic is quietly rebuilding its relationship with the Trump administration, and the government's appetite for its most powerful AI model appears to be winning out over politics. The detente underscores that the US doesn’t want to, and can’t afford to, sideline cutting-edge AI.

What happened: Just over a month after the Pentagon moved to block Anthropic’s models with “supply chain risk” concerns, signs of a truce have emerged. On Friday, Anthropic CEO Dario Amodei sat down with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent to discuss how federal agencies could access Mythos, the company's most advanced model, which Anthropic says poses an enormous cybersecurity threat. (So far, Mythos has only been released to a select group of organizations.) Both sides called the meeting productive. Meanwhile, the National Security Agency reportedly has access to Mythos Preview.

It’s not personal: The shift is largely due to necessity. Anthropic’s latest models are seen as best in class, particularly for cybersecurity. Walking away from that capability would mean giving up a competitive edge. According to reports from Axios, Wiles and Bessent both want the government to have a relationship with Anthropic. Bessent has reportedly separately encouraged major bank CEOs to test Mythos. And a US official told Axios that “every agency” except the Pentagon wants access to Anthropic models. There’s also a bigger strategic concern: Sidelining a major player like Anthropic could put the US at a disadvantage in the broader AI race.

So much for “corporate murder”: When the Pentagon fight started in February, some wondered if the supply chain risk designation would be fatal for Anthropic. Hardly. Not only did the company’s annual revenue run rate recently increase to $30 billion (up from $9 billion at the end of last year) thanks to soaring demand for its products, it’s also now valued around $800 billion (basically the same as OpenAI).

Bottom line: Agencies other than the Pentagon are likely to expand testing and deployment of Anthropic models, especially in areas like cybersecurity and intelligence analysis. The government will probably end up with a split relationship, suing the AI company with one hand and using its tools with the other. Meanwhile, Anthropic is starting to look like an AI zombie: The Pentagon tried to kill it, and it came back stronger. —AC

Smart glasses, for obvious reasons, keep getting banned in places like schools and courtrooms, but South Korean theaters are doing the opposite and handing them out at the door. As part of a pilot backed by the Korea Tourism Organization, theatergoers can now wear AI glasses that translate performances into their preferred language, displayed directly on the lenses (you can pick where, as well as the font size), per the New York Times. The idea is to reach global audiences without sacrificing the original language.

The glasses are camera-free (recording a live performance is, famously, frowned upon), which makes the whole thing feel a lot less dystopian than your average smart glasses deployment. And honestly, the use case makes sense: Imagine watching an opera without squinting at the tiny text on the seatback screen in front of you.

There are some caveats, though. One attendee told the NYT that wearing the AI glasses over his prescription pair was a little cumbersome—the glasses-over-glasses problem that anyone who's ever been handed chunky 3D glasses at a movie theater knows all too well. Gizmodo also notes that open-ear wireless earbuds paired with a translation app could do the same job with less hardware on your face. Still, it's a rare win for a tech category that usually makes headlines for all the wrong reasons. —SM

Chaos Brewing Meter: /5