The White House is reviewing AI with Anthropic while the Pentagon is still trying to squeeze it. Keep that in mind. Quick catch-up before we dig in: Anthropic says the Pentagon’s supply-chain-risk designation of Claude was retaliation for the company’s limits on mass domestic surveillance and fully autonomous weapons. That fight isn’t resolved. Meanwhile, the White House has separately talked with Anthropic CEO Dario Amodei about collaboration and protocols around the Claude Mythos model. This is Anthropic Pentagon Watch. Today, we start with the White House framework meeting, then the clock contractors are up against, and the Silicon Valley fight over who gets locked out of AI. First up—the meeting. This one's from Business Insider Africa:
On Tuesday, staffers from OpenAI, Anthropic, Google, and other tech companies convened in Washington to review the administration's new framework for reviewing advanced AI models. Now, the rest of the industry eagerly awaits answers on what will be included in the framework a structured way for President Donald Trump's administration to evaluate AI models' cyber capabilities.
At Tuesday’s White House meeting, OpenAI, Anthropic, and Google were in the room to discuss a cybersecurity review framework. Business Insider Africa says it’s aimed at certain advanced models before release—so it’s a process, not an open-ended pass to inspect every model. And the reported carveout matters: open-source models are outside the review plan, at least so far. Give closed labs a federal release gate while open weights skip it, and you’ve drawn a pretty convenient regulatory map for the biggest closed-model vendors. The administration calls this light-touch regulation; a pre-release federal cyber review feels more like a light touch with a clipboard. The meeting didn’t answer what triggers review, what the government can demand, or how companies challenge a finding. Here's Fellplaw:
Anthropic was notified by the United States Department of War (“DoW”) through a series of letters that it had been designated as presenting a supply chain risk in early March and that Anthropic products could no longer be used in certain defense-related work as a result. The federal government later clarified that the designation would become effective within 180 days to provide contractors time to transition away from covered Anthropic products and services.
The contractor fallout in Anthropic’s supply-chain-risk fight is coming into focus as the transition deadline approaches. Fellplaw puts it at 180 days, so primes eyeing FY2027 work are making vendor decisions while the lawsuits are still unresolved. And the clock is doing the punishing before a court settles the merits. If your defense-adjacent product was built around Claude, “transition away” means time to re-engineer, sign replacement contracts, and retest. Somebody gets that bill. Fellplaw flags 10 U.S.C. 3252 alongside FASCSA, and contractors shouldn’t blur them together. The Northern District of California injunction is still in force; the D.C. FASCSA case is still live after the emergency stay was denied. Legal clarity, in the government’s preferred two-court format. A cybersecurity walkthrough didn’t solve this. The procurement machinery has a 180-day fuse on it, and contractors are standing closest to it. Maximilian Schreiner, writing in THE DECODER:
Five of those weights said the options on the table included sanctions, a trade blacklist targeting Chinese makers of open models, and even a ban on US cloud companies doing business with those firms. Officials also discussed which models federal agencies should be allowed to use and whether the Commerce Department should impose new rules. After an outcry from Silicon Valley, the administration shifted course and is now focused on making American models more competitive, the NYT reports.
Kimi K3 is in the mix, and Anthropic and OpenAI are urging sanctions and a trade blacklist. Nvidia, Meta, Microsoft, and Google are pushing back. The labs want Washington to fence off open-weight competition; the platform companies want to keep selling chips, cloud, and tools. According to THE DECODER’s account of the Times reporting, the White House considered sanctions, a blacklist, and even limits on U.S. cloud business with Chinese model makers. The administration backed off for now after the industry fight, so none of those measures is a ban already in force. The timing is rich. Anthropic is facing a 180-day contractor transition clock while lobbying to restrict Kimi K3. Everybody’s suddenly very principled about supply-chain risk when the supply chain is somebody else’s model. The administration’s fallback is to make American models more competitive. Fine—then write policy that does it. When firms already selling competing products pitch a blacklist, give it a lot less benefit of the doubt. If you’re following Anthropic Pentagon Watch, check out The Data Center Daily—a daily briefing on how AI at scale is reshaping compute, hyperscaler capex, the power grid, semiconductor supply, and energy markets. Find it wherever you listen to podcasts.
Next, we’re watching the 180-day transition window for contractors moving away from covered Anthropic products—the next procurement checkpoint. Links to every story are in the show notes, so have a look at the ones you want to read more closely. That’s Anthropic Pentagon Watch for today. This is a Lantern Podcast.