Anthropic Restricts Claude Game-Cloning Use

Anthropic has reportedly updated the system prompt for Claude Code to stop the coding tool from being used to recreate existing video games. The change follows a surge of AI-assisted activity in retro gaming, including decompilations, recompilations, and ports that reproduce or closely derive from protected titles. Anthropic is not banning game development outright. Instead, it is drawing a boundary around using Claude to reconstruct existing works. The distinction is important. Developers often use AI coding assistants for original game logic, debugging, tooling, or prototypes, and those uses can remain permitted. The restriction targets a narrower problem: automated or semi-automated recreation of games that may infringe intellectual property. The update illustrates how quickly AI coding agents are running into established legal and commercial norms. A model that can inspect code, generate large amounts of it, and follow multi-step instructions may make copying or porting faster, even when the underlying legal questions are unchanged. For Anthropic, the move is also a governance signal. Model providers are increasingly shaping acceptable use through system prompts and product-level constraints, not only through terms of service. That can offer faster responses to emerging behavior, but it also raises questions about transparency and consistency. Users may need clearer guidance on where the line sits between inspiration, interoperability, and recreation. The broader lesson is that AI coding tools are becoming powerful enough to collide with IP norms in practical, visible ways. As a result, providers are likely to keep adjusting their products as they learn how customers use autonomous coding agents.