Anthropic appears to have ordered AI tool Claude to stop remaking video games
Anthropic says it updated Claude Opus 5.5 to steer the coding model away from recreating specific games.
TL;DR
- Anthropic confirmed an update intended to steer Claude away from recreating another developer’s specific game.
- Tom’s Hardware also reported user claims that the safeguards sometimes blocked broader reverse-engineering work.
- The company’s stated restriction concerns game cloning; its wider effect on legitimate analysis remains uncertain.
Anthropic updated Opus 5.5 to steer Claude Code away from recreating another developer’s specific game, Game Developer reported after the company confirmed the change. The update followed concerns about AI-assisted game cloning. [1] [2]
Tom’s Hardware reported user accounts that Claude Code or Opus 5.5 also refused some reverse-engineering requests users described as benign. These reports concern user experiences and do not establish how often the refusals occur. [1] [3]
The restriction is described as targeting recreation of specific games, rather than prohibiting all game development. Anthropic’s confirmation was narrower than some user reports about refusals during general analysis or security work. [1] [2] [3]
The published accounts do not quantify false refusals or explain the full enforcement criteria. It remains unclear how Anthropic will distinguish game cloning from authorized interoperability, preservation or defensive research. [2] [3]
Why it matters
The change shows how model safeguards can shape the boundary between creative generation and reproducing protected software. The unresolved question is how a rule aimed at cloning behaves in adjacent, legitimate technical work.
Editor's note
Anthropic confirmed a targeted adjustment; broader refusal reports are attributed to users and are not quantified.