The case centers around Anthropic’s attempt to prevent the federal government from using its AI, known as “Claude,” for autonomous weapons or mass surveillance of Americans.