What jumps out to me is not the existence of a safeguard, but how blunt it sounds. If the report is accurate, Claude Code is apparently treating “games” as a category it should just avoid, which is a very different thing from blocking obvious piracy. Those are not the same problem, and developers know it.
A blanket rule like that is exactly where useful work gets caught in the net. Rebuilding a board game for a client who owns the game, or working on a commissioned port, is not morally identical to ripping off someone else’s commercial IP. The source’s strongest detail is that the developer says Anthropic’s support bot told them the policy covers all customers, with no exception. If that’s really the rule, then Anthropic has chosen simplicity over nuance. That may be easier to enforce, but it’s also the kind of policy that makes a coding tool feel less like an assistant and more like a gatekeeper with a very fuzzy idea of what counts as legitimate work.
I also don’t think the timing is incidental just because Anthropic hasn’t announced anything. Silent server-side changes are exactly the sort of thing that make developers nervous, because they can’t tell whether they hit a bug, a policy experiment, or a permanent restriction. And once people start talking about memory or Git history carrying a refusal forward, you get into the unpleasant territory where the model is not just refusing a prompt, but refusing a project. That might be a misunderstanding, but it’s the kind of thing that deserves a clean explanation from Anthropic, not a support-bot shrug.
The part I’d actually want to see is a proper verification path. If Anthropic is worried about copyright abuse, fine, but then say so and give legitimate developers a way to prove they’re building under license or for the rights holder. Otherwise the policy just punishes the exact people who are least likely to be the problem.
Reference: Claude Code now refuses to help build copies of existing games
