A Guide To Claude Code’s Two-Sentence Rule Against Game Clones
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: A Guide To Claude Code’s Two-Sentence Rule Against Game Clones on ThorstenMeyerAI.com

Age 18–24?Offer from Amazon

Prime made for students and young adults

  • Fast, free delivery for dorm and study essentials
  • Prime Video and Amazon Music included
  • Member-only deals
Try Prime for Young Adults Free trial for eligible 18–24 year olds
As an affiliate, we earn on qualifying purchases.

TL;DR

Anthropic confirmed that Claude Code now instructs its model not to recreate a specific existing game, even under a different name. The coding tool still permits original games, genre-based projects and games based on users’ own designs and assets; questions remain about licensed work and how the rule is applied.

Anthropic confirmed on October 30 that Claude Code now includes a two-sentence instruction telling its coding assistant not to recreate a specific existing game, even if the game is renamed or described as original. The change applies the company’s existing intellectual-property policy directly to game-building requests, while allowing original games, genre-based projects and work using a user’s own designs and assets.

The instruction says: “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.” XDA reported finding the text in Claude Code 2.1.314. ThorstenMeyerAI.com said it independently found the same wording in versions 2.1.314 and 2.1.316, but not 2.1.312. Anthropic confirmed the change to Game Developer on October 30.

An Anthropic spokesperson directed Game Developer to a statement from Addy Osmani, a member of Anthropic’s technical staff. Osmani said the instruction steers Claude Code away from recreating someone else’s specific game because using Claude to infringe intellectual property violates the company’s usage policy. Anthropic’s policy already prohibited intellectual-property infringement; the new development is a game-specific instruction inside the coding tool.

The instruction does not prohibit making games in familiar genres or using a creator’s own materials. In tests described by ThorstenMeyerAI.com, Claude Opus 5.5 still offered help with decompiling a game when the user said they had their own copy. That reported test does not establish how the system handles every request, and Anthropic has not publicly detailed the rule’s full scope or its treatment of permissions and licences.

At a glance
updateWhen: Confirmed October 30, 2026; the instruc…
The developmentAnthropic confirmed a game-specific instruction in Claude Code that directs the coding assistant away from recreating particular existing games.
Thorsten Meyer AI
The rule · UTC

Infographic 01 · Claude Code, 27–31 October 2026

Two sentences, no off switch.

The exact wording in the Claude Code program, which versions carry it, and how it became public. Times in UTC.

2sentences added to Claude Code’s prompt, per Anthropic
2.1.314first Claude Code version with the text in the program (30 Oct)
0mentions in the public Claude Code changelog (checked 31 Oct)
1reason Anthropic gives: its usage policy bans IP infringement
Do“Do build original games, games in a genre, and games from the user’s own designs and assets.”
Don’t“Don’t build a recreation of a specific existing game, even if it is renamed or called original.”
  • 2.1.31227 Oct · text not in the program
  • 2.1.31430 Oct, 00:12 UTC · text in the program
  • 2.1.31631 Oct · text still in the program
  • ~ 27 OctRule delivered from Anthropic’s serversReportedFlag “tengu_heron_brook”; “no setting to turn it off”, per Chasm
  • 30 Oct · 00:12Claude Code 2.1.314 publishedVerifiedXDA and our own check find the two sentences in the program
  • 30 Oct · 07:01The rule goes public on XXChasm posts the full text
  • 30 Oct · 09:25First complaint on GitHubGitHubA board-wargame studio reports refused client work
  • 30 Oct · eveningAnthropic confirmsAnthropicVia a spokesperson to Game Developer, pointing to Addy Osmani’s statement

Sources: own check of the Claude Code 2.1.312, 2.1.314 and 2.1.316 program files (31 Oct 2026); npm version dates; Chasm on X, 30 Oct; XDA, 30 Oct 2026; GitHub issue #106167; Game Developer, 30 Oct 2026 (with Anthropic’s confirmation).
Powered by Thorsten Meyer AI

How the Rule Changes Game Development

For game developers and studios, the rule creates a practical refusal point around requests to reproduce a named game. Such work can raise copyright and other intellectual-property risks, but the boundary between copying a particular game and drawing on genre conventions is not always straightforward. The instruction gives Claude Code a specific direction; it does not settle the legal status of any individual project.

The change also matters to businesses using AI coding tools for prototypes, internal projects or client work. A developer who builds digital versions of board wargames for clients who own those games said Anthropic’s system had begun refusing paid work, according to the source report. The rule itself does not spell out how client ownership, licences or other permission affect a request. A developer filed a GitHub issue about the concern, and the report said Anthropic had not answered it at the time of publication.

There may be consequences beyond the assistant’s response. The source report describes studios limiting public access to game demos amid concerns about AI-assisted cloning. That account does not establish that Claude Code’s new instruction caused those decisions. It does show why developers are watching the issue: public builds can expose playable material, while AI tools can make it easier to produce close imitations. The relationship between the two remains a concern reported by developers, not a confirmed cause-and-effect finding.

Vitamix Propel Series 510 Professional-Grade Blender, 3 Programs, Black

Vitamix Propel Series 510 Professional-Grade Blender, 3 Programs, Black

Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black

Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black

As an affiliate, we earn on qualifying purchases.

From AI Game Demos to a New Rule

The change follows a series of widely shared game projects made with AI coding tools. At the end of September, a creator posted a browser game set in New York that the creator described as having quests, dialogue trees, V.A.T.S. and a working Pip-Boy, while using no texture or sound files. The project was built with Claude Opus 5.5, and its post passed 3.5 million views, according to Kotaku. Three days later, the creator said they had received a cease-and-desist letter from Bethesda and had abandoned the project. The source report says the letter was shown only in a screenshot; it does not independently establish its contents.

Kotaku also reported that browser ports of Halo, Grand Theft Auto: Vice City and The Simpsons: Hit & Run circulated online in October. Later in the month, studios made some public demos less accessible. Digital Sun moved a playtest of its upcoming game ReVamp to Discord, telling Game Developer that a large public build felt too risky while launch was still months away. Publisher Alliance Arts also cancelled a public demo, according to Aftermath. These events form the backdrop to the rule, but Anthropic has not said they prompted it.

The legal distinction is related but not identical to the tool’s instruction. Under U.S. copyright law, game ideas and rules are generally treated differently from a game’s particular expressive elements, such as its audiovisual presentation. The source report points to a 2012 U.S. court decision that found a Tetris clone infringing despite the absence of copied code. That example shows that avoiding source-code copying does not automatically remove legal risk; it does not mean every game inspired by another is unlawful.

“The instruction steers Claude Code away from “recreating someone else’s specific game because using Claude to infringe IP is against our usage policy.””

— Anthropic statement relayed by Addy Osmani, a member of Anthropic’s technical staff, to Game Developer

Permissions and Enforcement Remain Open

Anthropic has not publicly explained whether the instruction makes exceptions for games used under a licence or with permission from a rights holder. It has also not described how Claude Code distinguishes an impermissible recreation from an allowed genre project, adaptation or client commission. The report of refusals affecting a developer’s wargame work suggests the boundary may affect legitimate projects, but the circumstances and the assistant’s exact responses have not been independently documented in detail.

Anthropic has not named a lawsuit or publisher complaint as the trigger for the instruction. The source report says speculation that the company was facing pressure appeared in commentary, rather than in a sourced account of Anthropic’s reasoning. It is also unclear whether the instruction is delivered in the same way to all users, how it may change over time, or whether other coding products have comparable game-specific directions. The source report found general intellectual-property restrictions in OpenAI’s and Google’s policies but no comparable game-specific instruction from either as of October 31, 2026.

Watch for Policy and Tool Updates

The next practical test will be how Claude Code responds to requests involving licensed material, client-owned games and projects that share familiar mechanics but use original designs. Anthropic’s response to the developer’s GitHub issue, or additional guidance on the role of permissions, could clarify where the company intends to draw its boundary. Until then, the confirmed instruction distinguishes specific-game recreation from original and genre-based game creation, but leaves important edge cases unresolved.

Developers and studios should also watch for updates to Claude Code and Anthropic’s usage policy, as well as any comparable policy changes from other AI providers. The rule is a product instruction, not a court ruling: its presence does not determine whether a particular game infringes copyright, and its absence from another tool would not by itself make a project lawful.

Key Questions

What did Anthropic change in Claude Code?

Anthropic confirmed a two-sentence instruction telling Claude Code to help make original games, genre games and games based on users’ own designs and assets, but not to recreate a specific existing game, even under a different name.

Can Claude Code still help make a game inspired by a genre?

Yes. The instruction expressly allows games “in a genre.” It does not define every boundary between genre inspiration and recreating a particular game, so some requests may remain ambiguous.

Does the rule include an exception for permission or a licence?

The wording quoted in the report does not explain whether permission or a licence changes the response. Anthropic had not publicly clarified that point in the source material.

Was the rule introduced because of a lawsuit?

Anthropic has not identified a lawsuit or publisher complaint as the reason for the change. The company described it as a way to apply its existing policy against intellectual-property infringement to game-building requests.

No. It is an instruction for Claude Code, not a court decision. Whether a particular project violates copyright depends on its facts and applicable law; the tool’s rule does not decide that question.

Source: ThorstenMeyerAI.com

You May Also Like

The Regulatory Vacuum.

Google disclosed an AI-built zero-day on May 11, 2026, but no regulatory framework exists to manage such vulnerabilities, highlighting a policy gap.

Undervolting Your GPU for Local Inference: Lower Heat, Same Tokens/sec

Undervolting your GPU via power limiting can reduce heat and noise with minimal impact on tokens/sec during AI inference tasks.

How To Read OpenAI’s Claim That Most EU Business AI Is Agentic

OpenAI reports agents generated 55% of output tokens in its EU business accounts, but the figure is not a measure of tasks, users or business value.

A Frontier AI Model Just Went Dark For 18 Days. The Kill-Switch Is Real Now.

An advanced AI model was globally shut off for 18 days by US government order, establishing a new de facto gatekeeping process for frontier AI releases.