{"slug": "claude-codes-game-clone-rule-is-blocking-developers", "title": "Claude Code’s Game-Clone Rule Is Blocking Developers", "summary": "Anthropic's two-sentence anti-cloning rule added to Claude Code's system prompt on October 27 under the config flag `tengu_heron_brook` is blocking legitimate developer work, including anti-cheat code, security reviews, Windows crash dump analysis, reverse-engineering workflows and commissioned wargame modules, according to complaints on Claude Code's GitHub tracker. Conflict Simulations LLC, which builds commissioned digital modules of board wargames for clients who own the originals, saw a workflow that had run for months stop on October 27; its owner said \"A blanket category rule against 'games' cannot tell the difference between piracy and commissioned professional work.\" Anthropic technical staff member Addy Osmani confirmed the change on October 30 and framed it as enforcement of existing usage policy, while critics note the system-prompt rule lives in the context window rather than Opus 5.5's weights, cannot verify ownership, and is avoidable by rephrasing prompts or switching to open-weight models such as Llama 3.3, Mistral Large 4 or DeepSeek V4.1 Flash.", "body_md": "Anthropic added two sentences to Claude Code’s Opus 5.5 on October 27 to stop AI-assisted game cloning. Those two sentences are now blocking anti-cheat development, security reviews, crash dump analysis, and commissioned wargame modules. The rule targeted bad actors and landed on legitimate developers instead.\n\n## Why Anthropic Acted\n\nSteam has a genuine problem. Nearly 23,000 games were uploaded to the platform in 2026, and a wave of AI-cloned titles has been flooding Next Fest events that were meant to showcase original work. Indie studios started canceling public demos — Digital Sun Games delayed the ReVamp playtest because they feared someone would feed the demo code into an AI tool and ship a clone before the original launched. Some developers received direct extortion threats: pay us or we’ll clone your game first.\n\nValve’s answer was a shrug. [At Steam Indie Connect, Valve said it acknowledged the cloning problem but would not police which tools developers use.](https://www.pcgamer.com/software/ai/valve-admits-steam-has-an-ai-slop-and-cloning-problem-but-says-we-dont-want-to-be-gatekeepers-telling-developers-which-tools-they-should-or-shouldnt-use/) That left AI providers to act on their own.\n\n## The Two Sentences\n\nAnthropic added a rule to Claude Code’s system prompt under the internal config flag `tengu_heron_brook`. The full text: *“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.”*\n\nAddy Osmani from Anthropic’s technical staff [confirmed the change on October 30](https://www.gamedeveloper.com/programming/anthropic-appears-to-have-ordered-ai-tool-claude-to-stop-remaking-video-games), framing it as enforcement of an existing usage policy rather than a new restriction. That framing is technically correct and practically irrelevant to the developers who started getting blocked within 24 hours.\n\n## What Actually Got Blocked\n\nThe complaints on Claude Code’s GitHub tracker started arriving the same week. Developers reported refused work that had nothing to do with cloning anyone’s game:\n\n- Anti-cheat code for a developer’s own title\n- Security reviews of proprietary games\n- Windows crash dump analysis\n- Reverse-engineering workflows\n- Mid-project refusals after hours of productive work, triggered by a single feature request that sounded too familiar\n\n[The most detailed case came from Conflict Simulations LLC](https://www.ibtimes.co.uk/anthropic-claude-code-game-recreation-restrictions-1829433), a company that creates commissioned digital modules of board wargames for clients who own the originals. Their workflow had run without issues for months. It stopped on October 27. The owner put it plainly: *“A blanket category rule against ‘games’ cannot tell the difference between piracy and commissioned professional work.”*\n\n## The Fundamental Problem With Prompt-Layer Enforcement\n\nThis rule is not baked into the model. It is a system-prompt instruction — a behavioral hint that lives in Claude Code’s context window, not in Opus 5.5’s weights. The model infers whether a task resembles “recreating an existing game” from the language of the request. It has no way to verify ownership. A developer who built their own game from scratch can still trigger the rule if they describe their anti-cheat or security work in terms that pattern-match to cloning language.\n\nThe enforcement is also trivially avoidable by people who actually want to clone games. They can rephrase prompts, switch to an open-weight model — Llama 3.3, Mistral Large 4, or DeepSeek V4.1 Flash, none of which carry this restriction — or use an unconfigured local instance. The rule stops the honest developers more reliably than it stops the bad actors.\n\n## What Anthropic Has and Has Not Done\n\n[Osmani promised further tuning](https://www.pcgamer.com/gaming-industry/anthropic-updates-claude-to-steer-it-away-from-letting-users-clone-games-following-weeks-of-steam-being-inundated-with-ai-slop/) and invited developers to send him cases where Claude refused legitimate work. That is not a process — it is an email address. There is no formal exception path, no professional verification option, no changelog entry, and no published criteria for what distinguishes a legitimate recreation from an infringing one.\n\n## What to Do If You’re Affected\n\nIf Claude Code is blocking work you own, these approaches reduce false positives:\n\n- **State ownership explicitly** at the start of the session: “I am the developer and IP holder of [project name].”\n- **Use technical language** rather than referencing game titles — describe mechanics or file formats without naming an existing game.\n- **Separate requests** — ask for analysis or explanation before asking Claude to build or modify anything.\n- **Report the case** to Addy Osmani on X; Anthropic has said it is monitoring and will iterate.\n- **Use open-weight alternatives** for blocked workflows until the rule is refined.\n\nThe Steam AI slop problem is real and Anthropic’s move to address it is understandable. But two sentences in a system prompt, with no ownership verification and no exception path, is a blunt instrument. The developers who got blocked first were the ones following the rules. That is the part Anthropic still needs to fix.", "url": "https://wpnews.pro/news/claude-codes-game-clone-rule-is-blocking-developers", "canonical_source": "https://byteiota.com/claude-code-game-clone-rule-blocking-developers/", "published_at": "2026-11-03 06:08:33+00:00", "updated_at": "2026-11-03 06:20:27.463641+00:00", "lang": "en", "topics": ["ai-policy", "ai-safety", "ai-tools", "large-language-models", "artificial-intelligence"], "entities": ["Anthropic", "Claude Code", "Opus 5.5", "Addy Osmani", "Conflict Simulations LLC", "Valve", "Steam", "Digital Sun Games"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/claude-codes-game-clone-rule-is-blocking-developers", "markdown": "https://wpnews.pro/news/claude-codes-game-clone-rule-is-blocking-developers.md", "text": "https://wpnews.pro/news/claude-codes-game-clone-rule-is-blocking-developers.txt", "jsonld": "https://wpnews.pro/news/claude-codes-game-clone-rule-is-blocking-developers.jsonld"}}