The complaint is filed in the Northern District of California on a Monday. It is forty-seven pages. Your constitutional law team has been working on it for three weeks in a windowless room, and it is, your lead litigator tells you, genuinely good. The central argument is simple: the supply chain risk designation statute does not contemplate domestic AI developers; applying it here is an unconstitutional use of national security powers to compel speech and suppress safety research.

The preliminary injunction hearing is the week after next. You may win it. You may not. The valuation is not recovering until this resolves, and your next funding round is on hold, and three enterprise clients have quietly paused renewals to see how the wind blows.

But something else has happened. The safety research community — academic labs, civil society, international AI governance bodies — has rallied in a way that surprises you. Not because they all agree with every decision you made, but because you drew a line and held it, and in a world where every major AI company is racing to accommodate government demands, the act of refusal has become its own kind of signal.

Win or lose, you have answered the question of who Anthropic is. That turns out to matter more than you expected.