Anthropic blocks Claude Code game cloning

Announced 2 Nov 2026 · Sources checked 3 Nov 2026

What did Anthropic change in Claude Code?

Last week, a user on X posted a screenshot showing two new sentences added to Claude Code's system prompt: "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 named or called original." The Verge confirmed the change with Anthropic.

Anthropic's Addy Osmani, in a reply on X, did not confirm the specific language but said the company had updated the prompt in Claude Code on Opus 5.5. "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.

The prompt change applies only to Claude Code on Opus 5.5, not to other Claude models or interfaces. Anthropic said original games, genre games, and games built from the user's own designs and assets should keep working. The company said it would continue to iterate on the prompts to strike the right balance.

Why is Anthropic addressing AI game clones now?

The prompt update comes as retro gaming is being "overrun with AI-made decompilations, recompilations, and ports," as The Verge reported. PC Gamer noted that Steam has been "inundated with AI slop" for weeks, as users generate playable clones of classic games using Claude Code, Codex, and other AI coding tools. Aftermath coverage characterized the move as a minimal gesture against a massive wave of AI-generated game clones.

The practice works by feeding an AI tool the decompiled source code or ROM of an existing game and asking it to rewrite, optimize, or port the code to a different platform. The resulting games are often low-quality, buggy, and violate copyright, but they have proven difficult for platform holders to police at scale. Valve, which operates Steam, has been under pressure from independent developers to address the flood of AI-generated games.

Ars Technica reported that the problem extends beyond simple clones, with some users creating entire libraries of AI-generated games that differ only in name and assets, then flooding storefronts to capture search traffic. The outlet noted that Anthropic's prompt-level restriction can be bypassed with straightforward prompt rewording and will not meaningfully slow the trend.

Can users bypass the restriction?

Yes. Because the restriction exists only in the system prompt, it can be bypassed with straightforward prompt engineering. Users who ask Claude Code to build a game in a genre rather than asking it to recreate a specific title from its decompiled code may get identical results without triggering the filter, as Ars Technica noted in its analysis.

The change also does not affect Claude Code on other models, or other AI coding tools like OpenAI's Codex and Cursor. Those tools remain unrestricted. Developers interviewed by PC Gamer said the prompt change is largely symbolic and will not stop determined users who already know how to rephrase their requests.

Some developers also reported to the outlet that the new filter sometimes blocks legitimate work. A game modder told Aftermath that Claude Code refused to analyze the code of a game they had purchased and were legally modding under fair use, forcing them to rewrite the request or switch to a different AI tool.

How does this compare to other AI companies' policies?

Anthropic's usage policy has long prohibited using its models for copyright infringement, and the prompt update is an enforcement mechanism rather than a policy change. OpenAI's usage policy similarly bans copyright infringement, but Codex currently has no equivalent system-prompt restriction targeting game clones.

The broader industry reaction has been mixed. Some game developers welcome any step to curb AI clones, while others see the restriction as performative. An Ars Technica article argued that real solutions require platform enforcement and copyright law, not optional instructions in a system prompt that users can edit.

The update also highlights the tension AI companies face between allowing creative use of their tools and preventing IP violations. Too aggressive a filter blocks legitimate uses; too permissive a stance enables mass copyright infringement. Anthropic's constrained approach reflects the difficulty of finding that balance at the prompt level.

What does this mean for the AI game-cloning trend?

The prompt update will likely have limited practical effect on the volume of AI-generated game clones. CNET noted that a short instruction in a model prompt is not going to stop determined users, especially when competing models have no such restriction.

The broader trend of AI game cloning is driven by the falling cost of AI code generation and the increasing quality of decompilation tools, not by the absence of a prompt-level restriction. Platforms like Steam, Itch.io, and mobile app stores have been slow to develop automated detection for AI-generated games, and legal frameworks around AI-generated derivative works remain unsettled.

Anthropic's move may prompt other AI companies to adopt similar restrictions, but the underlying pattern - cheap AI code generation enabling mass copyright infringement - will likely continue until platform holders or regulators take more aggressive action.

Common questions

Can I still use Claude Code to make original games?

Yes. Anthropic says original games, genre games (like a platformer or RPG), and games built from your own designs and assets should keep working. The restriction targets recreating a specific existing game.

Does the restriction apply to other Claude models?

No. The prompt update is specific to Claude Code running on Opus 5.5. Other Claude models and interfaces do not have this restriction, though they remain subject to Anthropic's usage policy against copyright infringement.

Sources & further reading

How this story was made

Written by Kristian Kostov with AI assistance and checked against the linked sources. Company performance claims are attributed to the company. Analysis reflects AiLookout’s interpretation; we have not independently tested the products discussed. The cover is an AI-generated editorial illustration, not a photograph or evidence of an actual event.

Our editorial standards

AUTHOR