The White House has not publicly released the voluntary framework for testing frontier AI models that it finalized in August.
Page A2From§Eachthe supper edition — 10 September 2026
War Department loses suit for calling Anthropic a security risk, staffs office gutting safeguards for tech contracts.
As it ran on the front
The word illegal shows up exactly once in what I've got written down, and I'm going to get to it, but first I want to read you the whole list, in order, and then I'm going to stop talking for a second.
Here's the list. On August 28th, a federal judge found that the government had no legal basis for branding Anthropic a national security risk. On September 10th — thirteen days later — Anthropic put out its own report saying that over the past eight months, it disrupted people trying to use its AI models for research into biological weapons. That same report also describes stopping attempts to build drone swarms; it doesn't say for how long that one's been going on, just that it happened. A Pentagon office staffed by people who used to work at the big tech companies is taking apart AI safeguards, while those same companies compete for the military contracts that come after the safeguards are gone. The White House finished a framework back in August for testing whether these AI models are safe to use. It hasn't put that framework out where the public can read it. And this week, a poll came back saying two out of three voters want Congress to ban the kind of AI the government keeps loosening the rules on. That's the whole list. I'm not adding to it.…
…(cont) I went out to the garage after supper with that list in my head and pulled the toolbox down — the one with the drawer that sticks, you've got to hit it with the heel of your hand to get it open. I laid every tool out on the bench in the order I found them. Wrench, wrench, the ratchet, the socket set that's missing the 10-millimeter, the level that hasn't read level since the Carter administration. I didn't skip the ones I didn't like the look of. That's the whole demonstration. You don't get to reach into a drawer, pull out the one tool you want to call dangerous, and leave the rest of the drawer alone.
Here's what I keep circling back to. The ruling on the branding came down August 28th. The report on the bioweapons and the drone swarms came out September 10th. In between those two dates, did the government put out the safety-testing framework it finished in August? Not that I can find. Does the Pentagon office written up this week hire people who used to draw a paycheck from the same tech companies now bidding on the contracts that follow once the safeguards are gone? According to what I've got, yes. Is the White House that ordered Anthropic blacklisted the same White House sitting on the only paperwork that would tell you whether any AI — including the military's — is safe to run? Same address, is all I can tell you.
And two out of three voters this week said they'd back a bill banning the kind of AI the government keeps deregulating. I don't have a question about that number. I've just got the number.
Sorry — one more, and then I'm done. The ruling says the government had no legal basis for calling Anthropic a risk. The Pentagon office taking the safeguards apart so those contracts can go through — is anybody calling that illegal?
“They spent months insisting Anthropic was too dangerous to touch, and a judge just said there was no legal basis for that label at all. Meanwhile there's a Pentagon shop full of ex-tech guys stripping the actual safety checks off AI headed for military contracts, and the one piece of paper that would tell you if any of it's safe is sitting in a drawer in the White House. Two-thirds of the country wants this stuff banned outright, and the only thing getting banned around here is the company that said no to the military.”
“Look, national security moves fast, and Anthropic tied the military's hands on autonomous weapons — somebody had to step in, that's just good governance. The office at the Pentagon isn't gutting safeguards, it's modernizing procurement so we're not out-innovated by — did I say modernizing? That's not what I — fine, it's modernizing. And the testing framework isn't hidden, it's finished, which is a totally different thing from released, and — you know what, next question.”
The receipts
A little-known Pentagon office staffed by Big Tech alums is dismantling safeguards and accelerating AI warfare as their former employers compete for lucrative military contracts.
As whistleblowers in the artificial intelligence industry speak out about the extraordinary risk posed by their technology if it remains unchecked, a new poll out Wednesday reveals that more than two-thirds of US voters agree with a new proposal in Congress to pause artificial intelligence development until guardrails are set up and ban…
As industry insiders issue louder warnings about the catastrophic danger of unchecked artificial intelligence, Anthropic has revealed in a report out Thursday that it had to "disrupt" several attempts by scientists to use its AI models for research that could contribute to the creation of biological weapons. In the report , titled…
‘Emerging threats’
· Judge rules Trump administration illegally punished Anthropic from the morgue, 28 Aug 2026
· Federal judge rules that War Department’s designation of Anthropic as a national security risk was illegal from the morgue, 28 Aug 2026
· Markey Urges Congress to Block Trump-Saudi Deal, a 'Starting Gun for a Middle East Nuclear Arms Race' from the morgue, 26 Aug 2026
This page is a back-issue: the story as it ran, receipts as they were. The current edition is at the front.