Anthropic's Claude AI blocks video game cloning and reverse-engineering

What it claims

Tom's Hardware (Bruno Ferreira) reports that Claude Code and Claude Opus 5.5 began refusing game-recreation and some reverse-engineering requests in the last week of October 2026. Twitter user Chasm posted the alleged rule text in full: "Do build original games, games in a genre, and games from the user's own designs and assets. Don't build a recreation of a specific existing game, even if it is renamed or called original." The rule permits genre games and work from the user's own assets. It blocks recreating a named title, and renaming the result does not get around it.

The article ties the timing to a separate policy. Anthropic's Cyber Verification Program, announced October 6, 2026, limited its frontier models' advanced cybersecurity capabilities to entities covered by Project Glasswing, effective immediately. The author suggests one ruleset may cause both the game and the security refusals. He notes that refusal reports began only about 24 days after that announcement.

Evidence

The evidence is user reports, not tests. Claude Code's GitHub repository has a run of issues about refusals in several situations:

• halfway through an existing game project,
• when replicating a game mechanic,
• when patching a tabletop-like game.

The author says at least one of these reporters is an established developer. Reddit and other platforms carry reports of refusals for analysing a Windows crash dump, removing junkware from a Samsung TV, and using the Reverse Engineer Anything (REA) tool at all. Self-described game developers say Claude refused anti-cheat development and security reviews of their own games, citing triggered safeguards. The article also reports that OpenAI deactivated the account of a user cleaning up TV software.

The article gives two motives. One is legal exposure to software and game studios. The other is a wave of AI-made clones on Steam Next Fest, plus extortion scams that threaten to clone small studios' demos before release.

Limits

Anthropic has made no statement, and the rule text is unverified. Commenter Ayzore says the change is a system-prompt edit that targets only prompts like "remake X game in Y style," and that reverse engineering, modding and decompilation remain allowed. That commenter calls the headline misinformation. Commenter salgado18 argues the rule would still block an original infinite runner that starts from an existing game and diverges from it.

Commenters suggest Chinese, open-weight or local LLMs as alternatives for reverse-engineering work. The article does not reproduce any refusal or measure how often refusals happen.