Pentagon Wanted Anthropic's AI to Surveil Americans' Data
Failed Pentagon-Anthropic negotiations reportedly centered on using Claude to analyze bulk data collected on US citizens.
The Pentagon's doomed relationship with Anthropic just got a lot more interesting. According to a source who spoke to The Atlantic's Ross Andersen, the Defense Department pushed right up until the end to use Anthropic's AI for analyzing bulk data collected about Americans.
The talks ultimately collapsed when Pete Hegseth moved to terminate the government's relationship with the AI company entirely. But the revelation about what the Pentagon actually wanted is the real story here.
This wasn't about battlefield logistics or foreign intelligence analysis. The government wanted to point one of the world's most capable AI systems at datasets about its own citizens. Anthropic, a company that has built its brand around AI safety and responsible deployment, apparently couldn't get comfortable with that arrangement.
The failed deal highlights the growing tension between national security agencies hungry for AI capabilities and the companies building them.