Anthropic adds rules to prevent Claude from cloning existing video games

The new system prompt

Anthropic recently updated the underlying instructions for its Claude models to address concerns about game cloning. The change was first identified by a social media user and AI enthusiast who noticed a new rule added to the Claude Code prompt around October 27. This specific instruction is contained within a file named "tengu_heron_brook" in the system's configuration.

The new rule explicitly tells the AI to build original games, games within a specific genre, or games based on a user's own designs and assets. However, it strictly forbids the AI from building a recreation of a specific existing game. This applies even if the user attempts to rename the game or claim the output is an original work.

Independent verification of this change came from XDA, which confirmed the rule appeared in version 2.1.314 of the Claude Code binary released on October 30. The update suggests that Anthropic is actively monitoring how its tools are used to generate content that mimics existing commercial products.

Why Anthropic is acting

The company's primary motivation for these new rules appears to be the prevention of intellectual property infringement. Anthropic software engineer Addy Osmani confirmed on social media that the company added these two sentences to steer the model away from recreating specific games. He noted that using Claude to infringe on someone else's IP is a violation of the company's usage policy.

By placing these restrictions directly into the system prompt, Anthropic creates a layer of defense against potential legal liability. If a user asks the AI to build a clone of a popular title, the model is now instructed to refuse the request. This helps the company maintain a safer environment for its users while attempting to mitigate risks associated with copyright law.

The update is part of a broader trend where AI providers are trying to establish clear boundaries for their tools. While the AI remains highly capable of generating code and mechanics, these prompts serve as a guardrail. They ensure that the technology is used to create new content rather than simply duplicating what already exists in the market.

Technical implementation details

The way these rules are integrated into the AI's behavior is quite specific. XDA reported that the "tengu_heron_brook" rule had been used for various other purposes in earlier versions of the software. This indicates that Anthropic has a history of using specific prompt bundles to manage model behavior and safety.

Furthermore, the Claude Opus 5.5 model appears to have its own distinct and independent instructions regarding game recreations. These instructions are bundled into the system prompt, which provides the AI with its core identity and behavioral guidelines. This multi-layered approach ensures that the restrictions are consistently applied across different model sizes and versions.

The goal of these prompts is to differentiate between creative assistance and direct copying. The AI is still encouraged to help with

Impact on the cloning trend

While Anthropic's new system prompt serves as a formal barrier against direct game cloning, it may not be enough to halt the current trend. The massive wave of AI-assisted game porting and copying that has emerged over recent months is already well underway. Experts suggest that while these rules help shield the company from some liability, they are unlikely to slow the broader momentum of the technology being used to replicate existing titles.

The move highlights a growing tension between the capabilities of large language models and the protection of intellectual property. Even with these new restrictions in place, the underlying technology remains powerful enough to facilitate complex tasks. This creates a situation where the AI's ability to generate functional code might still outpace the effectiveness of simple text-based instructions designed to prevent specific outcomes.

Detection and verification

The discovery of these new rules began with a social media user and AI decompilation enthusiast known as "chasm." This individual was among the first to notice the change late last week, noting that a new rule appeared to be added to the Claude Code prompt around October 27. This proactive monitoring by the community shows how quickly users can identify shifts in how AI models are being steered by their developers.

XDA later independently verified these findings, confirming that the new rule is present in version 2.1.314 of the Claude Code binary released on October 30. Their investigation also revealed that the "tengu_heron_brook" rule had been utilized for different purposes in previous versions. Additionally, the Claude Opus 5.5 model seems to have its own unique and independent instructions for restricting game recreations within its system prompt bundle.