New Delhi [India], October 31: Anthropic has confirmed it added a restriction to Claude Code designed to stop users from recreating specific existing video games, following a wave of AI-assisted clones that has unsettled independent developers and prompted at least one studio to cancel a public playtest. The company says the measure targets intellectual-property infringement, not game development generally.
The change was first noticed by an X user known as Chasm, who reported around October 27 that Claude Code had received a server-side rule under the internal flag “tengu_heron_brook.” The instruction 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 said there was no setting to disable it.
ANTHROPIC IS CRACKING DOWN ON VIDEO GAME REWRITES
Claude Code picked up a new rule around Oct 27, sent from Anthropic’s servers under a flag called “tengu_heron_brook”. There’s no setting to turn it off.
The full rule: “Do build original games, games in a genre, and games from the user’s own designs and assets. Don’t build a recreation of a specific existing game, even if it is renamed or called original.”
Is this the end of game rewrites, modding and everything in between?
— chasm (@chasmmmmmmmmmmm) October 30, 2026
Anthropic confirms the change
An Anthropic spokesperson directed reporters to a statement from Addy Osmani, 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 on X.
Osmani said original games, games within a genre and games built from a user’s own designs and assets “are meant to keep working,” adding that Anthropic would “continue to iterate to ensure we’re striking the right balance.”
The restriction is narrower than a blanket ban on AI-assisted development. It is aimed at requests to reproduce a particular title, even when the project is renamed or described as original. Reports say the instruction appeared in Claude Code version 2.1.314 and may be applied remotely through Anthropic’s server-side configuration system.
Developers cancel public tests
The update lands as some studios reconsider how much of an unreleased game they are willing to expose. Digital Sun, the Valencia-based developer behind Moonlighter, said it had cancelled a planned public Steam playtest for its upcoming “Tower Defensevania” ReVamp because it feared a public build could be reverse-engineered, cloned and released before the original game.
“With launch still months away, a large public build of ReVamp felt like too big a risk,” the studio said. It moved to a smaller, invitation-based Discord playtest beginning October 29, with a public playtest now expected around January. ReVamp is scheduled to launch on March 25, 2027.
Digital Sun’s statement also highlighted the wider stakes for independent studios, which often rely on public demos and playtests to build an audience. “Showing games early is how indies get seen and find the fans who keep us going, so we hope platforms keep strengthening protections for original work,” the studio said.
Valve points to discovery, not bans
The pressure has also reached Valve. At Steam Indie Connect in South Korea on October 27, Steam business team members Tom Giardino and Marc Wilkinson said Valve does not intend to ban AI-assisted games from Steam. Instead, Valve said it would focus on improving discovery and recommendations.
On clones specifically, Giardino acknowledged that bad actors have used early trailers and demos to rush out AI-made copies before the original ships, but said the practice predated generative AI and has accelerated. He argued developers should reduce copyability by continuing to add features, personality and creative choices rooted in their own perspective.
Limits of a prompt rule
The policy may prove difficult to enforce consistently. Anthropic’s instruction is a prompt-level steer rather than a technical block, and reports note that users determined to make clones could rephrase requests, use other AI tools, or turn to open-weight models without comparable restrictions.
There are also early signs of overblocking. A GitHub issue from a developer who said they build licensed board-wargame adaptations reported that Claude refused work on projects the developer described as properly licensed. The cause and scope of that refusal remain unclear.
For now, Anthropic’s move is one of the clearest attempts by a major AI company to draw a line between helping people make games and helping them copy someone else’s. Whether it protects developers—or merely pushes cloning to less restricted tools—will depend on how consistently the rule is enforced and whether storefronts such as Steam build stronger safeguards for original work.
