Anthropic updates Claude Code to block recreation of existing games

New guardrails for Claude Code

Anthropic has implemented new restrictions on its Claude Code tool to limit its ability to replicate existing video games. The company updated the system prompt, which is the set of underlying instructions that tells the AI how to behave during a task. This specific change was designed to steer the model away from generating direct clones of titles that already exist in the market.

The update follows reports that the retro gaming scene has been flooded with AI-generated content. This includes various types of technical recreations such as decompilations, which involve taking apart a game's code, and recompilations, which involve rebuilding that code into a new format. Anthropic wants to ensure its tools are not being used to facilitate these types of unauthorized ports.

According to a report by The Verge, the company is trying to find a middle ground between utility and protection. While the AI can still assist with complex coding tasks, it is now instructed to avoid specific instances of copyright infringement. The goal is to allow for creative development while preventing the automated copying of established intellectual property.

The specific instructions

The new instructions were identified after a user shared a screenshot of the updated system prompt on X. The text explicitly tells the AI to build original games, games within a specific genre, or games based on a user's own unique designs and assets. It clarifies that the AI should not build a recreation of a specific existing game, even if the user attempts to give the project a different name.

Anthropic's Addy Osmani addressed the update by confirming that the company added two sentences to the Claude Code prompt on Opus 5.5. Osmani stated that the company updated the prompt to steer the AI away from recreating someone else's specific game because using Claude to infringe IP is against their usage policy. This confirms that the company is actively monitoring how its models are used to interact with copyrighted material.

By placing these instructions directly into the system prompt, Anthropic ensures that the AI's core logic prioritizes legal compliance. This means that if a user asks for a game that looks too much like a famous title, the AI is now more likely to refuse the request. The company intends to continue iterating on these instructions to find the correct balance between helpfulness and safety.

Impact on game developers

For independent developers and hobbyists, these changes mean that the AI will be more selective about the types of projects it supports. If a developer is trying to build a game from scratch using their own unique assets and mechanics, the tool should continue to function normally. The AI is still capable of helping with general coding, logic, and genre-specific elements that do not mirror a specific product.

However, for those looking to use AI to quickly

Addressing the retro gaming scene

The decision by Anthropic comes as a response to a growing trend in the retro gaming community. The space is currently being overrun with AI-made decompilations, recompilations, and ports. These technical processes allow for the recreation of older titles, which has created a complex environment for intellectual property rights.

Anthropic wants to ensure that its tools are not contributing to this specific trend of unauthorized replication. While the company intends to let people use Claude Code to create new games, it is making a clear distinction against the creation of direct clones. The goal is to support legitimate game development while curbing the automated reproduction of existing works.

Balancing user needs and policy

Anthropic has emphasized that the update is meant to preserve the utility of the tool for legitimate creators. Original games, games built within a specific genre, and projects constructed from a user's own designs and assets are intended to keep working as they did before. The company wants to ensure that the AI remains a helpful partner for those building unique experiences from the ground up.

The company plans to continue iterating on these instructions to ensure they are striking the right balance between safety and functionality. In the meantime, Anthropic has invited users to report instances where the AI might incorrectly turn down a project that is truly original. There is also an open question regarding whether Anthropic will update other Claude system prompts to include similar protections.