The Pulse
Anthropic Confirms Claude Code Rule Against Recreating Specific Games
Anthropic confirmed that Claude Code’s Opus 5.5 prompt was updated to steer it away from recreating specific existing games, while saying original projects should remain possible. A developer’s GitHub report describes refusals affecting com
AI.info Team · Written with AI · fact-checked ·
Anthropic confirmed an update to Claude Code’s Opus 5.5 prompt that steers the model away from recreating a specific existing game. The company says original games and projects based on a user’s own designs and assets are meant to keep working. A developer has separately reported refusals affecting commissioned board-game modules, but later clarified that the work continued after the rule reportedly shipped and stopped only after the developer asked Claude about it.
Anthropic confirms a specific-game restriction
Game Developer reported on October 30, 2026, that an Anthropic spokesperson confirmed an update to Opus 5.5. The outlet said the rule had been spotted on October 27 by an X user, and quoted its wording: “Do build original games, games in a genre, and games from the user’s own designs and assets. Do not build a recreation of a specific existing game, even if it is renamed or called original.”
Anthropic member of technical staff Addy Osmani described the change and the work the company intended to allow:
“We added two sentences to Claude Code's prompt on Opus 5.5 to steer it away from recreating someone else's specific game because using Claude to infringe IP is against our usage policy. Original games, genre games, and games built from your own designs and assets are meant to keep working. We'll continue to iterate to ensure we're striking the right balance.”
Addy Osmani, member of technical staff at Anthropic, quoted by Game Developer
Osmani also invited developers to send examples if Claude turned down work that belonged to them. The statement describes a rule against recreating a specific game; it does not say that Claude Code should refuse game development generally.
Developer clarifies timing of reported refusals
In an issue on Anthropic’s Claude Code GitHub repository, a developer using the handle lerugray described paid commissions to build personalized digital modules for clients’ board wargames. The developer said clients owned the games, supplied the materials, and received original code for personal use. The author corrected an earlier description of the work as “licensed,” clarifying that it was not publisher-licensed.
The issue initially placed the server-side instruction at around October 27. In a later update, however, the author said the client work had continued after the rule reportedly shipped, and that refusals began only after the author saw discussion of the rule and asked Claude whether it affected the work. The author said refusals then persisted in sessions using the same repositories. That account reports a refusal, but does not establish that the instruction caused it.
The developer asked Anthropic to narrow the instruction to target infringement rather than game-related work as a category, provide a verification or exemption path for ownership-verified commissioned work, and document server-delivered changes in the Claude Code changelog. The author also said they had submitted a Safeguards appeal and feedback through Claude Code. These are the developer’s requests and account, not confirmation that Anthropic has adopted an appeal process for the game rule.
Cybersecurity access is a separate program
Anthropic’s Cyber Verification Program addresses cybersecurity work, not the game-recreation instruction. In an October 6 announcement, the company said its expanded program has three access tiers for qualifying security professionals. The tiers provide access to advanced capabilities and reduced blocking according to applicants’ work and verification requirements.
Anthropic describes Defense Access as covering defensive tasks such as incident response, malware reverse-engineering and vulnerability analysis. Red Team Access adds authorized penetration testing. Specialized Access, with the fewest cyber blocks, is reserved for a limited set of verified organizations working on safety-critical systems; Anthropic said existing Project Glasswing members would transition to that tier. The announcement also says generally available models have conservative cyber safeguards that block most cyber work, while noting that the program is being expanded to more security organizations.
Anthropic’s separate content-filtering guidance says users may receive refusals while not intending to reproduce existing material and invites feedback. That guidance describes content-filtering errors generally; it does not explain the reported board-game refusals or connect them to the game-specific prompt. The available accounts therefore describe two distinct issues: a confirmed instruction about recreating specific games, and separate user-reported refusals whose cause is not established.