Here's a tense situation brewing in Washington: the Pentagon has reportedly drawn a line in the sand for AI company Anthropic. According to reports, Secretary of War Pete Hegseth has warned the company that it could be kicked out of the Defense Department's supply chain if it doesn't agree by this Friday to let its technology be used in all lawful military applications.
Think of it as a classic government contractor dilemma, but with cutting-edge artificial intelligence. Hegseth called Anthropic's CEO, Dario Amodei, to Washington for a meeting on Tuesday—a meeting the Pentagon had previously confirmed. The discussions were, by all accounts, tense.
The Secretary of War's reported ultimatum comes with a big stick: the threat to invoke the Defense Production Act (DPA). This Cold War-era law allows the president to essentially commandeer domestic industry for national defense. If invoked here, it would let the Pentagon use Anthropic's AI tools even without the company's consent. Both former President Donald Trump and President Joe Biden have used this same act before, notably to address medical supply shortages during the pandemic.
So, what's the big disagreement about? It boils down to what Anthropic calls its "red lines." The AI startup has built certain ethical safeguards into its terms of service, specifically refusing to allow its technology to be used for mass surveillance of American citizens or for developing autonomous weapons that operate without a human "in the loop." The company argues its current AI systems aren't reliable enough for such lethal, high-stakes roles and has pushed for new safeguards.
The Department of War, however, sees these restrictions as overly limiting. From the Pentagon's perspective, having an AI partner that won't work on certain lawful applications is a problem. It's a fundamental clash between a company's ethical guardrails and a government agency's desire for unrestricted access to powerful tools.
The stakes for Anthropic are high. Being removed from the Pentagon's supply chain is a serious move, one typically reserved for companies linked to foreign adversaries. Such an action would also send a clear signal: Anthropic's AI is considered critical to current Pentagon operations. Sources indicate the move could also lead to legal repercussions for the company.
Perhaps most tellingly, the Pentagon isn't putting all its eggs in one basket. Reports indicate Hegseth is already in discussions with other major players in the AI space as potential alternatives. Those talks include Google (GOOGL), Google (GOOG), OpenAI, and Elon Musk's xAI. It's a reminder that in the race for military AI, there are other capable contenders waiting in the wings if one company balks at the terms.
Meanwhile, Amodei has been publicly vocal about his broader concerns. In a podcast last week, he expressed discomfort with the rapid concentration of AI power and wealth among just a handful of companies. He also warned that AI advancement is like an approaching "tsunami," whose impact many are underestimating. Those comments add an interesting layer of context to a negotiation that's fundamentally about the control and application of that very power.
The Department of War and Anthropic did not immediately respond to requests for comment on the reported ultimatum. All eyes are now on Friday's deadline to see if one side blinks, or if the Pentagon starts seriously shifting its AI strategy elsewhere.
