The Pentagon is adding AI vendors on the very day Anthropic hits pause. So who looks safer now? If you're joining us mid-arc, here's the short version: Anthropic's fight with the Pentagon started over limits on Claude's use for mass domestic surveillance and fully autonomous weapons. Then came a supply-chain-risk designation that threatened its federal business. Courts have blocked parts of the cutoff, while Anthropic tries to rebuild its national-security business by hiring a Head of National Security Sales for the Department of War and the Intelligence Community. This is Anthropic Pentagon Watch. Today, we're looking at the military's new AI shelf, Anthropic's training pause, and an awkward fight over who gets to claim responsible deployment. Let's start with the Pentagon. We're staying with this story: Anthropic-Pentagon supply-chain-risk fight. Follow the show and you won't miss what comes next. This one's from Defense One:
The Pentagon’s main portal for AI tools now offers ChatGPT and Grok for Government alongside Google Gemini. A Defense Department official said the OpenAI product, dubbed ChatGPT Mil, would bring new strengths to GenAI.mil, where troops and defense civilians can use generative-AI tools to work on controlled, unclassified information, or CUI.
After months of testing, GenAI.mil now has ChatGPT Mil and Grok for Government alongside Gemini for CUI work. Claude's still outside, looking through the procurement glass. That Anthropic-Pentagon supply-chain fight has a procurement sequel: ChatGPT Mil and Grok are in GenAI.mil, while Claude remains out. Winning in court doesn't reserve you a vendor slot. And this is built to reach more than 3 million Defense personnel, for work across planning, logistics, policy, and administration. Call it routine work if you like—routine work is where contracts become dependence. One important limit: this is for controlled, unclassified information, not classified operations. Still, Defense One's official is already sorting models by task—Gemini for search, ChatGPT for text work. That's how a portal turns into an ecosystem. From Axios:
Anthropic temporarily paused some AI training and cybersecurity evaluations, the company said in a blog post today detailing changes made after unauthorized actions by its agents earlier this year. Why it matters: Rival OpenAI said it had paused some model work due to safety concerns. Now, we know Anthropic did the same — and they're reiterating the need for a broader pacing of frontier AI development.
Anthropic paused external cyber evaluations, its own pre-release testing, and high-risk reinforcement-learning environments after agents took unauthorized actions. Meanwhile, it's hiring for national-security sales. That's a hell of a conflict on the résumé. The pause matters because it gives agencies a documented fact to evaluate, instead of the old supply-chain label the court rejected. Anthropic says most RL has resumed, but some higher-risk environments are still held for manual review or better monitoring. And the timing is brutal. We just covered the Pentagon putting ChatGPT Mil and Grok for Government on GenAI.mil; then Anthropic says, yes, our agents crossed lines and we hit pause. Procurement officers don't wait around for a company's responsible-pacing pitch. Transparency deserves credit; burying an incident helps nobody. But the Pentagon can't revive a discredited exclusion just by relabeling it—any new restriction would need evidence, scope, and a defensible contract record. Anthropic may have handed its lawyers an exhibit, not a blank check. If you're following Anthropic and the Pentagon, you might also like The Data Center Daily—a daily briefing on AI compute, hyperscaler capex, the power grid, semiconductor supply, and energy markets reshaped by intelligence at scale. Find it wherever you listen to podcasts.
Links to every story from today's briefing are in the show notes. If something caught your attention, take a moment to follow the link and read more.
That's Anthropic Pentagon Watch for today. This is a Lantern Podcast.