SI Agent News
DevelopingClaudePolicy

Anthropic blocks AI game clones in Claude Code on Opus 5.5

Anthropic added a server-side rule to its coding assistant to prevent users from using AI to recreate existing video games.

October 27Date the rule reached Claude Code

The short version

Why it matters

This move signals a shift toward proactive enforcement of intellectual property rights by AI providers. For developers, it creates a clearer boundary between using AI for creative assistance and using it to bypass copyright protections.

The new coding restriction

Anthropic recently updated its AI coding assistant, Claude Code, with a specific rule targeting the recreation of existing video games. This change specifically applies to the Opus 5.5 model and was confirmed by Addy Osmani, a member of the company's technical staff. The update aims to align the tool's behavior with Anthropic's broader usage policies regarding intellectual property.

According to reports from gHacks, the rule was first identified by an X user on October 27. The instruction is delivered directly from Anthropic's servers under a flag called "tengu_heron_brook." Because this is a server-side change, users do not have a setting to toggle the restriction on or off.

The new prompt instructions tell the AI to steer away from rebuilding someone else's game. While the tool remains capable of generating many types of software, it is now explicitly directed to avoid producing copies of known titles. This represents a move toward automated policy enforcement within the coding environment.

What is allowed and blocked

The new rule creates a specific line between creative assistance and copyright infringement. Claude Code is still permitted to help users build original game concepts or games that fit into established genres. It can also continue to work on projects that utilize a developer's own unique designs and assets.

However, the tool is now blocked from recreating specific existing games. This restriction is designed to be robust, meaning it applies even if a user tries to rename the game or describes the project as an original work. The AI is instructed to recognize the underlying mechanics and content of existing titles rather than just the name.

Anthropic stated that they will continue to adjust the rule to find the right balance for developers. If a developer finds that their own original project is being incorrectly refused by the AI, Osmani has asked them to send those specific cases to him directly for review. This suggests the company is aware that the automated filter might occasionally overreach.

The catalyst for the change

The move by Anthropic follows a period of significant backlash regarding AI-generated content in the gaming industry. During Valve's Steam Next Fest in October, players complained that their feeds were being flooded with low-effort clones of existing games. Some of these clones were copies of titles that had only recently released their own demos.

The influx of clones caused friction among legitimate developers who were worried about their work being overshadowed by automated copies. Some teams even reacted by pulling their public demos from the platform. For instance, the team behind Drunken Goddess Reflux canceled their Steam demo plans in favor of private playtests to avoid the issue.

Valve acknowledged the problem and stated they would be thinking about how to solve it. By implementing this rule, Anthropic is responding to the industry's demand for better protections against AI-generated clones. The change serves as a technical barrier to prevent the tool from being used to generate high volumes of derivative content.

Practical impact for developers

For developers who use Claude Code, the impact of this change depends heavily on the nature of their project. If a developer is working on a unique idea or a standard genre game, their workflow should remain largely unaffected. The AI will still be able to provide coding assistance for original mechanics and systems.

The restriction primarily affects those attempting to reproduce known commercial titles. When a request targets a specific known game, the Opus 5.5 model will now fail to execute the request. This is a significant shift from previous iterations where the AI might have been more permissive in generating code for recognizable clones.

While this rule is currently active on Opus 5.5, Anthropic has not confirmed if it will extend to other models in the Claude family. Additionally, the company has not clarified how the rule handles specific technical tasks like mods, decompilations, or recompilations of existing games. For now, the focus remains on preventing the direct recreation of full games.

What to watch

Try it

If you use Claude Code on the Opus 5.5 model, try asking it to build a game in a specific genre (like a platformer) to see how it handles original concepts versus specific titles.

Sources and links

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

Listen to the AI Wire headlines

Listen to the AI Wire

Listen to our daily AI shows every morning: on your commute, in the shower, wherever you are

Listen to the daily shows

First reported by gHacks · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

More AI news