Anthropic has updated the system prompt for Claude Code and Claude Opus 5.5 to prohibit the recreation of specific existing games. The restriction reportedly applies even when developers give the resulting project a different name.
The company introduced the change amid growing concern in the games industry over AI-generated copies. Existing ideas and publicly shared demos can also be repurposed with AI assistance before a game’s release, potentially creating pressure and even extortion attempts, according to the report.
Anthropic said its models may still assist with game development when users provide their own designs and assets. The company cited the risk of intellectual-property violations behind the policy change. Anthropic employee Addy Osmani said on X that recreating existing games violates the company’s usage rules, while the restriction may also help reduce the risk of copyright litigation.
However, the stricter safeguards appear to be affecting some legitimate projects. Claude is reportedly blocking certain attempts to analyze games through Reverse engineering, with the widely used Reverse Engineer Anything tool seemingly among the projects affected.
Users have also reported on GitHub that the AI refuses to help reproduce game mechanics or patch game code. Complaints on Reddit and other platforms reportedly include developers in unrelated fields, such as people trying to investigate a Windows crash or remove junkware from a smart TV.
It remains unclear whether all reported cases involve legitimate uses. Some complaints, however, have reportedly come from established and credible developers. Anthropic has said it plans to refine the rules to find a better balance. Osmani invited users to report cases in which Claude rejects work that belongs to them.
Sources: t3n.