Anthropic has introduced new restrictions to Claude Code designed to prevent its AI coding assistant from recreating existing video games, but the changes are already raising concerns among developers who claim their legitimate projects are being blocked. First identified by developer Chasm on X, the new instructions reportedly appeared around October 27, 2026, through a server delivered configuration named tengu_heron_brook. The rule directs Claude Code to avoid recreating specific existing games, even when renamed or presented as original creations, while continuing to support independent development, genre inspired projects, and games built using original assets.
Anthropic's Addy Osmani confirmed the change, explaining that 2 sentences were added to the Claude Opus 5.5 instructions to discourage intellectual property infringement. The company emphasized that it is not prohibiting AI assisted game development and intends to continue supporting original projects.
"Original games, genre games, and games built from your own designs and assets are meant to keep working."
— Quote by: Addy Osmani, Anthropic
However, the implementation is already creating complications. In an official Claude Code GitHub issue filed on October 30, a developer representing Conflict Simulations LLC reported that the assistant had started refusing to work on personalized digital board game modules commissioned by paying customers. The developer explained that these projects use original code and customer supplied materials, with the finished modules intended for personal use by customers who own the corresponding physical games. Although the projects are not officially licensed by their original publishers, the developer argues that the new restrictions are incorrectly interfering with previously functional development workflows.
The complaint also revealed an unusual behavior: Claude Code reportedly continued working on the projects after the new restrictions were introduced, but began refusing requests once the developer explicitly asked whether the rules applied to their work. Additional reports have emerged involving reverse engineering, game modification, and troubleshooting activities, although there is insufficient evidence to establish that every refusal originates from the same game specific restriction. The updated instructions do not explicitly prohibit legitimate modding or game optimization.
The development comes amid growing concerns about AI generated game clones appearing on platforms such as Steam, where independent developers have increasingly questioned how their original work can be protected from rapid imitation. As recent discussions about AI in game development highlight, automated coding tools are becoming increasingly important to modern production workflows. Anthropic has acknowledged the possibility of unintended restrictions and encouraged affected developers to report incorrectly rejected projects while it continues refining the system.
Protecting original game developers from unauthorized AI generated copies is a reasonable objective, particularly as coding assistants become capable of recreating increasingly complex gameplay systems. However, applying restrictions without reliably identifying ownership, licensing, or legitimate development intent risks disrupting the same creative communities these tools are supposed to empower. Modding, preservation, accessibility improvements, and independent development frequently involve working with existing game mechanics and software. Anthropic's challenge will be establishing safeguards that discourage infringement without unnecessarily restricting legitimate technical work or undermining developer confidence in AI assisted programming.
Should AI coding assistants prevent users from recreating existing games, or should developers have greater freedom to use these tools for modding, preservation, and personal projects?
