Anthropic's AI model Claude is rejecting requests to duplicate existing video games, even when users apply new names or claim the work is original, based on multiple user reports. Developers also state that these new restrictions block routine troubleshooting and reverse engineering tasks that do not involve copying games.
Tom's Hardware compiled these reports, which involve Claude Code, the AI coding tool from Anthropic, and its Opus 5.5 model. Programmers encountered these rejections over the past week, and Anthropic has not released any official statement about the situation.
What Claude Is Now Refusing to Do
According to X user Chasm, Claude operates under a guideline stating: "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."
No official confirmation exists verifying this exact rule text. However, the GitHub page for Claude Code features numerous complaints from developers whose tasks were blocked midway through development, while copying a game mechanic, or while patching a tabletop-style game. At least one complaint originates from a developer with an established background.
Put simply, prompting Claude to create "a racing game" or "a game like a platformer" should function properly. Asking the tool to reconstruct a specific, famous game under a different name faces blocking.
Legitimate Users Are Getting Blocked Too
The broader complaint centers on reverse engineering, which involves taking apart existing software to comprehend its operation. Users across Reddit and other websites report that Claude rejected tasks such as analyzing a simple Windows crash dump, removing junk software from a Samsung TV, or using the Reverse Engineer Anything tool.
Certain game developers also state that Claude declined assistance with anti-cheat development and security reviews for their personal games, indicating that safety safeguards were activated.
This behavior might connect to the Cyber Verification Program launched by Anthropic on October 6. Under this framework, access to advanced cybersecurity features of frontier models restricts to organizations covered by Project Glasswing. Even so, rejection reports began appearing roughly a week ago, more than three weeks after that program started.
Disagreement exists regarding the accuracy of reverse-engineering claims. A reader on Tom's Hardware contended that the adjustment only halts "lazy cloning," such as prompts asking Claude to rebuild a specific game in another style, while modding and decompilation work continues to function.
Why Anthropic Is Doing This
Legal protection serves as the probable motive. AI corporations face risks of lawsuits from software developers and game studios when tools assist in copying or cracking products. AI-generated clones of software packages like Adobe applications and Microsoft Office, alongside numerous console game recreations, heighten this concern.
This issue also impacts small studios. Steam Next Fest recently drew criticism for containing numerous AI-generated games copied from current titles, and certain developers reportedly faced extortion scams threatening to clone game demos into complete games ahead of original releases.
Anthropic operates alongside other companies taking similar actions. OpenAI recently deactivated an account belonging to a user attempting to clean up television software.
What This Means for You
Users employing Claude to construct games can still expect original ideas, genre titles, and projects using personal designs and assets to function. Conversely, requests resembling copies of famous games, even with altered names, may face rejection.
Developers, freelancers, and students using Claude Code for debugging, security testing, or software analysis may encounter increased rejections, even for benign tasks. Online communities already recommend Chinese, open-weight, or locally run AI models as substitutes for such tasks.
For paying subscribers of Claude, certain dependent tasks may fail, making it prudent to test regular workflows prior to relying on the tool for client projects. This situation unfolds as Anthropic addresses user scrutiny regarding cache resets that increase costs, alongside the company filing for an initial public offering.
Critics highlight the irony that AI companies trained models on extensive copyrighted internet content while now restricting users due to copyright worries.
