Anthropic adds two sentences to stop Claude Code from cloning existing games

Anthropic added two sentences to Claude Code's Opus 5.5 prompt to prevent the model from recreating specific existing video games. The restriction addresses a surge in game clones built with AI. X user Chasm identified the instruction on October 27. The restriction permits original games, games within a genre and projects using assets supplied by the user. However, the restriction prohibits reproducing a particular existing game, including under a different name.

Addy Osmani, a member of Anthropic's technical staff, confirmed the modification on X on October 30. Osmani pointed to Anthropic's policy against intellectual property infringement.

Developers say the restriction also prevents legitimate work beyond game cloning. Claude Code's GitHub issue tracker contains accounts of refusals during routine game development. Developers report blocked debugging on projects the developers own. Other accounts describe refusals involving anti-cheat development, security reviews and a basic Windows crash dump. Anthropic has not disclosed how the model determines ownership of a project. Osmani pledged additional adjustments and asked developers to report refusals involving work the developers own.

Modders have used AI to put Mario into Elden Ring and bring Vice City to a web browser. Mashable reported that an AI-assisted PlayStation 5 jailbreak enabled Microsoft's Halo 2 to run on Sony hardware. Some developers say scammers threatened to turn the developers' demos into complete competing games before the original games launch. Digital Sun Games cancelled a public playtest of ReVamp because of concerns about copies.

The restriction precedes Anthropic's Nasdaq debut. Anthropic has already paid $1.5 billion to resolve claims concerning pirated books used for training. Anthropic's IPO prospectus, filed on October 26, identifies potential liability arising from misuse by third parties as a risk. The prospectus records the settlement with authors as an expense for 2025. Investors are awaiting Anthropic's IPO pricing before assigning valuations to competing AI laboratories.

Developers on Reddit and X say the developers will move to models from China, open-weight models or locally run models. Lower-cost competitors, including OpenAI's GPT-6.1 Terra, are already competing through pricing. Mashable described the competitive dilemma: "...the least ethical AI might be poised to become the most popular."