Anthropic has added a rule to Claude Code that tells its coding agent not to rebuild a specific existing game, after weeks in which developers have watched AI-assisted copies of their work turn up on Steam.
The change was first spotted by X user Chasm, who said an update to Claude Code's "rules" was pushed sometime on October 27, Game Developer reported. The rule is listed under the name "tengu_heron_brook" and reads: "Do build original games, games in a genre, and games from the user's own designs and assets. Do not build a recreation of a specific existing game, even if it is renamed or called original." Chasm reports there is no way to switch it off, according to Game Developer.
What Anthropic said
Anthropic's confirmation came in a post on X, which we couldn't check directly, so the wording here is as Game Developer and PC Gamer quote it. Game Developer says an Anthropic spokesperson pointed it to a statement from Addy Osmani, whom both outlets describe as a member of Anthropic's technical staff.
"We added two sentences to Claude Code's prompt on Opus 5.5 to steer it away from recreating someone else's specific game because using Claude to infringe IP is against our usage policy," Osmani wrote, as quoted by Game Developer. He added that "original games, genre games, and games built from your own designs and assets are meant to keep working," that Anthropic would "continue to iterate to ensure we're striking the right balance," and: "In the meantime, if Claude turns down something that's yours, send it my way."
By Anthropic's own account, then, the change is two sentences in the prompt the model works under. That matters for what comes next.
Why now
PC Gamer, which picked up Chasm's find via Game Developer, ties the change to multiple reports from developers whose games and demos were copied with the help of AI, which it says led several studios to cancel planned public demos for fear the code would be reverse engineered and cloned. Game Developer names one: Digital Sun Games cancelled the upcoming public playtest for its next game, ReVamp.
PC Gamer also reports that the most recent Steam Next Fest showcased twice as many demos as the previous event, with AI-assisted games accounting for most of that increase. Valve has acknowledged the problem without banning AI tools; at Steam's Indie Connect event, Valve's Tom Giardino said the company doesn't want "to be gatekeepers telling developers which tools they should or shouldn't use," PC Gamer reports.
Tom's Hardware adds a darker detail: some game developers have reportedly been threatened with having their demos cloned into finished games before they can release the originals.
The catch for developers
The rule isn't landing cleanly. Tom's Hardware reports that Claude Code's GitHub has seen a run of issues from users complaining of refusals on ordinary game work, including partway through development, when replicating a game mechanic and when patching a tabletop-style game. It allows that some of those reports may come from users with bad intentions, but says at least one comes from a developer with a real track record, and that separate user reports describe refusals on general reverse-engineering tasks, though it is unclear whether those are linked to the game rule.
Game Developer is sceptical that the rule will stop determined copycats. It points out that safety limits on AI tools have repeatedly been sidestepped with rhetorical tricks, that it's unclear how the software could tell whether a user owns a game idea or its assets, and that users replying to Chasm noted cloners can switch to open-weight models without such limits.
Mason's take: Two sentences in a prompt won't fix Steam's clone problem on their own, and Anthropic's talk of iterating suggests it knows that. The interesting part is who moved. Valve, which runs the store, has said it doesn't want to tell developers which tools to use; the toolmaker has now drawn a line itself. That line is blurry, though. "A recreation of a specific existing game" is easy to say and hard to judge from inside a coding session, and the early refusals suggest some of the people paying for the vagueness so far are developers working on their own games. If Anthropic wants this to hold up, the "send it my way" escape hatch has to turn into something more dependable than one employee's inbox.
Reporting notes