Anthropic Files Declarations to Contest Pentagon Risk
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic filed two sworn declarations and a reply brief in a California federal court ahead of a March 24 hearing before Judge Rita Lin.
- Sarah Heck, Anthropic’s Head of Policy, declared that the Pentagon’s claim Anthropic demanded an approval role over military operations is false and never raised in negotiations.
- Sarah Heck also pointed out that the Pentagon’s concern about Anthropic disabling or altering its technology mid‑operation first appeared in the government’s court filings, not in prior talks.
- Sarah Heck attached a March 4 email from Under Secretary Emil Michael stating the two sides were “very close” on autonomous weapons and mass‑surveillance positions, contradicting the later risk designation.
- Thiyagu Ramasamy, Anthropic’s Head of Public Sector, asserted that Claude models deployed in air‑gapped, government‑secured systems cannot be remotely disabled, updated, or accessed by Anthropic.
- Thiyagu Ramasamy further said Anthropic employees have U.S. security‑clearance vetting, making Anthropic the only AI firm with cleared personnel building models for classified environments.
Why it matters: Anthropic’s challenge could preserve its ability to negotiate AI contracts without a blanket national‑security label, while the Pentagon risks losing a tool to block technologies it deems unsafe. The case also tests whether AI‑safety positions are protected speech, potentially shaping future government‑industry risk assessments.
