Something happened in Washington this week that most people scrolled right past.
On June 2, President Trump signed an executive order on AI. Same day, Anthropic expanded Project Glasswing — their restricted cybersecurity initiative built around Claude Mythos — from 50 organizations to 150, across more than 15 countries. Power companies. Hospitals. Water systems. Communications infrastructure.
Two things on the same day. Worth noting.
A glasswing butterfly is nearly invisible until the light hits it right. So are the security flaws Mythos was built to find — and so, apparently, was Anthropic's leverage.
Because four months ago, this same company was labeled a "supply chain risk to national security" by the Pentagon.
Here's what that designation actually meant: The Department of Defense wanted to use Claude for autonomous weapons systems and mass surveillance. Anthropic said no. The Pentagon reached for a statute designed for foreign adversaries and pointed it at an American company.
That wasn't a vulnerability story. That was a values story.
Anthropic held the line. They sued the federal government in two courts. They kept Mythos restricted even when the pressure to expand came from the top. And they kept building.
When Mythos launched in early April, Anthropic gave access to roughly 50 organizations — security firms, government agencies, critical infrastructure operators — and asked them to use it for one purpose: find the holes in your own systems before someone else does.
Those partners found more than 10,000 high or critical-severity vulnerabilities. In weeks.
That's not a product demo. That's a national security capability. And Washington noticed.
Anthropic CEO Dario Amodei met with White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent. Trump told CNBC they had "very good talks" and that Anthropic is "shaping up."
Translation: we need what they have.
The executive order that followed didn't mandate compliance. No forced licensing. No pre-approval requirements. It created a voluntary framework — frontier AI labs can give the government early access to new models, up to 30 days before public release, for cybersecurity evaluation. With IP protections. With confidentiality agreements.
A framework Anthropic could work with. And they signed on.
The 150 new organizations joining Glasswing this week aren't tech companies experimenting with a shiny tool. They're power utilities. Water systems. Healthcare networks. Hardware manufacturers. The people responsible for keeping the lights on.
Anthropic committed up to $100 million in usage credits and $4 million to open-source security organizations to make it happen. That's not a press announcement. That's a company putting real money behind the conviction that defense beats reaction — every time.
You are not a power grid. You're probably not joining Project Glasswing.
But here's what this week confirmed, loudly and in writing: AI is infrastructure now. Not a trend. Not a productivity experiment. The same category as electricity and broadband.
When that's true, the question isn't whether to use it. That question is already settled. The real question is whether you understand the environment well enough to make smart decisions about where it fits — and whether you have someone helping you think it through, or whether you're just reacting to whatever lands in your inbox next.
The companies that will do well in the next five years aren't the ones with the most tools. They're the ones who knew what the tools were actually for.
The glasswing was there the whole time. This week, the light just hit it right.
Mollie Barnett is the founder of State & Signal AI Systems and writes The Messenger on AI, technology, and the forces reshaping how we work. She is based in East Northport, NY.
Then the glasswing showed its wings.