How a Contractor’s Claude Agent Got Into the Pentagon

Judgment

Anthropic's deployment-policy leads and the contractor's security officers now have to decide whether an agentic coding tool can keep standing credentials into a classified DoD system, trading developer speed for a clear chain of accountability when the agent, not a cleared employee, holds the access.

Basis

On WIRED's Uncanny Valley, Zoë Schiffer introduced the show's scoop about "how a defense contractor plugged Anthropic's Claude into a secure Department of Defense system," and Leah Feiger called it "one of the most outlandish uses of AI in the government we've seen so far." The mechanism the framing implies: a contractor wired Claude Code to a personal AI agent and gave that agent reach into a secure Pentagon system, so the entity exercising access is software acting on an employee's behalf rather than a named, cleared person. That shifts the control point — the vendor's usage policy governs the model, but the contractor's own credentialing and logging govern what the agent can touch, and the two regimes are not built to talk to each other. The episode pairs this with Anthropic's IPO filing, where host Brian Barrett noted the financials show "a real business," with revenue of $16.3 billion in the first half of 2026 and a valuation moving from $183 billion to $965 billion — scale that makes any deployment-limit question a market question, not just a compliance one.