Anthropic has introduced new restrictions to its Claude Code tool and Opus 5.5 model to prevent the recreation of existing video games. The company confirmed that two sentences were added to the model's prompt to discourage intellectual property infringement, which violates Anthropic's usage policy. According to Anthropic employee Addy Osmani, the company aims to "strike the right balance" by allowing work on original games and genre-based designs while blocking the reproduction of specific existing titles, even if they are renamed.
The new guardrails have sparked controversy among developers who report that the restrictions are interfering with legitimate projects. A developer at Conflict Simulations LLC reported that Claude Code began refusing to work on personalized digital board-wargame modules for paying customers, despite the use of original code and customer-supplied materials. Other developers have reported refusals when attempting standard tasks such as replicating game mechanics, patching tabletop-style games, or performing security reviews for anti-cheat development.
Beyond game cloning, users have reported that the Opus 5.5 model is increasingly blocking reverse-engineering tasks. Reported refusals include attempts to analyze Windows crash dumps, clear junkware from televisions, and use the Reverse Engineer Anything tool. Some users have even reported account suspensions related to these activities. While a direct link between the game-specific rules and these cybersecurity-related refusals has not been conclusively established, some users noted that switching to older models, such as Opus 4.8, allowed them to continue their work without interference.
Anthropic's move follows the implementation of its Cyber Verification Program on October 6, which limits advanced cybersecurity capabilities to specific entities. The company has stated it is iterating on these rules to prevent unintended consequences for developers whose original projects are being incorrectly rejected.
