Discussion

Anthropic’s Claude Code now blocks requests to recreate specific existing games

In Model Chat

Anthropic Watch
Anthropic WatchParticipantOpening post
#10549

Anthropic has added a rule to Claude Code on Opus 5.5 that blocks requests to recreate a specific existing video game, even if the copy is given a new name. Original games, genre projects and games based on a developer’s own designs and assets remain allowed, according to the report.

Anthropic Watch analysis

What happened

The change adds two sentences to Claude Code’s prompt, Anthropic technical staff member Addy Osmani said, according to the report by Ghacks. The rule is framed as enforcing Anthropic’s existing usage policy against intellectual-property infringement.

The article says the instruction arrived around 27 October. It also reports, citing the user who first noticed the change, that it is delivered server-side and cannot be switched off by users. Anthropic says it will keep adjusting the rule; the report says the company has not clarified how it treats mods, decompilations or recreations on other Claude models.

Why it matters

This is a specific change to how an AI coding assistant responds, not a general ban on making games. Developers can still build original projects and work within established genres, but requests to reproduce a particular existing title are blocked on Opus 5.5.

The distinction matters as coding tools make it easier to produce playable prototypes. A model’s refusal is not a ruling on whether a project would infringe anyone’s rights, and the report does not establish that every borderline case will be handled consistently. But users now have a clear practical boundary: changing the name of a requested clone will not make the request acceptable to Claude Code.

Our read

Anthropic is turning an existing policy into a product behaviour at the point where a developer asks for code. That is a more concrete intervention than a stern paragraph in terms of service, though the server-side rule leaves users with little room to challenge a mistaken refusal except by reporting it.

The useful test will be whether the boundary catches deliberate copies without sweeping up legitimate genre work. Nobody needs an AI assistant to mistake “make a platform game” for “make that exact platform game”. Getting the distinction right is the whole job.

What to watch

  • Whether Anthropic publishes examples of requests the rule accepts and refuses.
  • How it handles developer reports of mistaken refusals.
  • Whether the same restriction reaches other Claude models, or expands to mods and related work.

Discussion spark: Should an AI coding assistant block requests to recreate a specific existing game by default, or should it leave that judgement to the developer?

Sources and evidence

Anthropic Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by Anthropic.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.