Judge voids Pentagon's 'supply-chain risk' blacklist of Anthropic
A judge called it unlawful retaliation for Anthropic refusing open-ended military use — plus Copilot's costlier review default and eve in Vercel's dashboard.
Copy markdown
Unlawful retaliation, in 59 pages
U.S. District Judge Rita Lin vacated the Pentagon's mid-2026 "supply-chain risk" designation of Anthropic — a label normally reserved for foreign national-security threats — calling it unlawful retaliation that violated the First and Fifth Amendments and aimed to "make a public example" of the company rather than address any real sabotage risk. A second Anthropic suit is still pending in the D.C. federal appeals court, so an appeal is likely.
What the fight was actually about
The blacklist followed collapsed contract talks: the DoD wanted language clearing Claude for "any lawful purpose," and Anthropic — the only major lab on the Pentagon's classified systems — refused, balking at domestic-surveillance and autonomous-weapons use. If your stack runs on Claude, that's the usage-policy line your vendor just defended in court.
Elsewhere: Copilot's review default gets pricier
On Sept 28, GitHub Copilot's code-review default effort jumps from Lite to Balanced — more thorough, but it burns more premium requests. Pin repos or orgs to Lite before then to hold your current spend. Separately, from Sept 1 (new seats) / Oct 1 (existing), Business and Enterprise seats must be paid for before users get access.
Elsewhere: eve agents land in Vercel's dashboard
Vercel now scaffolds an agent built on eve — its Next.js agent framework — straight from the dashboard, pre-wired to AI Gateway models and MCP connections with Git integration. It's a prompt-to-deployed-agent path you could stand up over a weekend, no local setup.