Anthropic has confirmed that it quietly updated Claude Code, its terminal-based coding agent, to stop users from recreating existing video games — a change that surfaced just as the games industry scrambles to respond to a wave of AI-built clones of popular and upcoming titles.

The restriction was first spotted by a developer known as Chasm on X, who noticed that a new rule had been pushed into Claude Code on October 27. The rule, listed under the internal label "tengu_heron_brook," tells the model: "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," according to Game Developer. Chasm reported that there is no way to disable the rule. For continuous coverage of how AI toolmakers are responding to the clone wave, track the latest AI developments as they unfold.

Anthropic Confirms the Change

An Anthropic spokesperson referred questions to a statement posted on X by Addy Osmani, a member of technical staff at the company. "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. "Original games, genre games, and games built from your own designs and assets are meant to keep working. We'll continue to iterate to ensure we're striking the right balance."

Osmani also extended an invitation to developers who get caught in the filter: "In the meantime, if Claude turns down something that's yours, send it my way."

Why Now: The AI Clone Wave

The timing is not a coincidence. In recent weeks, fear has spread among independent developers that fast-follow clones built with AI coding tools could reach the market before the games they copy. Spanish studio Digital Sun Games canceled the public playtest for its upcoming game ReVamp specifically because it feared someone would feed the demo into an AI tool and ship their own version first, Game Developer reported.

Developers have also pressed Valve to explain how it prevents AI-generated clones from supplanting original games on Steam, prompting the company to offer a rare explanation to developers in South Korea earlier this week. Against that backdrop, Anthropic's move reads as an attempt to keep its flagship coding tool out of the IP-infringement supply chain.

Legitimate Work Is Getting Caught Too

The problem, according to multiple reports, is that the restriction is catching work that has nothing to do with cloning. Claude Code's GitHub repository has seen an influx of issues complaining about refusals during standard game development — including work that stopped halfway through a project, attempts to replicate a single game mechanic, and patches for a tabletop-style game, as documented by Tom's Hardware.

Users also claim the model is blocking innocuous reverse-engineering tasks, including jobs run through the popular Reverse Engineer Anything (REA) tool, analyzing a Windows crash dump, and clearing junkware from a Samsung TV. Some purported developers say Claude refused to help with their own games' anti-cheat development and security reviews. The refusals follow Anthropic's Cyber Verification Program, introduced October 6, which gated access to the frontier models' advanced cybersecurity capabilities behind its Project Glasswing initiative.

Will Two Sentences Actually Stop Clones?

Skeptics argue the rule will only deter the most low-effort copycats. Safety limits on generative AI tools have repeatedly been sidestepped with rhetorical tricks, and it is unclear how the software could ever verify whether a user genuinely owns a game concept or its assets. As many users responding to Chasm pointed out, if Claude becomes too restrictive for clone-making, vibe-coders can simply switch to open-weight models that carry no such restrictions.

There is also a business tension embedded in the change: Anthropic actively courts developers with Claude Code subscriptions, and every refusal that lands on a legitimate project is a support ticket, a GitHub issue, and potentially a churned customer. Osmani's offer to review false positives suggests the company is aware the filter will misfire.

For now, the rule stands as one of the first times a major AI lab has hard-coded a content-category restriction into a developer tool's default behavior in response to a specific wave of misuse — and a test of whether prompt-level rules can hold the line where usage policies alone have not.

For game developers, the practical guidance is unchanged from what Anthropic has said publicly: building original games, working within a genre, and creating from your own designs and assets should keep working as before. Developers who believe Claude is wrongly refusing their own work now have a named escalation path in Osmani's offer, alongside the standing option of filing an issue on the Claude Code repository. Whether that is enough to keep legitimate game work flowing — while the tools simultaneously close the door on cloners — will show up quickly in the GitHub issue tracker, where the verdict is already being written one refusal at a time.

---

Stay Ahead of AI

Get the latest AI news, analysis, and breakthroughs — all in one place.

Read more AI news →