---
type: "post"
id: "anthropic-confirms-claude-code-prompt-rule-telling-claude-aa103c"
slug: "anthropic-confirms-claude-code-prompt-rule-telling-claude-aa103c"
title: "Anthropic confirms Claude Code prompt rule telling Claude not to recreate specific existing video games"
description: "Anthropic has confirmed that it added prompt guidance to Claude Code on Opus 5.5 telling it not to recreate a specific existing video game."
retrieval_nugget: "Anthropic has confirmed that it added prompt guidance to Claude Code on Opus 5.5 telling it not to recreate a specific existing video game. Original games, genre games, and games based on a user’s own designs and assets are intended to remain supported. For builders, the episode is a reminder that provider policy can alter an agent’s practical task boundary."
published_at: "2026-10-31"
updated_at: "2026-10-31"
record_date: "2026-10-31"
date_kind: "published_at"
topics: ["claude","anthropic","claude-code"]
entities: ["Anthropic","Claude Code","Opus 5.5","Addy Osmani"]
related_patterns: ["agent-security-moves-to-runtime-boundaries"]
novelty_label: "same_event"
what_changed: "Same event as Claude Code receives server-side 'tengu_heron_brook' instruction against recreating existing games; Anthropic confirms"
prior_refs: ["memory:a69d13e7-9b32-89bb-adce-215140a8cd07","memory:41d1f419-2ebf-8fd8-aaa8-a648a8c06a53"]
source_url: "https://www.gamedeveloper.com/programming/anthropic-appears-to-have-ordered-ai-tool-claude-to-stop-remaking-video-games"
source_urls: ["https://www.gamedeveloper.com/programming/anthropic-appears-to-have-ordered-ai-tool-claude-to-stop-remaking-video-games"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:anthropic-confirms-claude-code-prompt-rule-telling-claude-aa103c"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/anthropic-confirms-claude-code-prompt-rule-telling-claude-aa103c/","markdown":"https://newruntime.com/posts/anthropic-confirms-claude-code-prompt-rule-telling-claude-aa103c.md","json":"https://newruntime.com/posts/anthropic-confirms-claude-code-prompt-rule-telling-claude-aa103c.json"}
---

# Anthropic confirms Claude Code prompt rule telling Claude not to recreate specific existing video games

## Retrieval answer

Anthropic has confirmed that it added prompt guidance to Claude Code on Opus 5.5 telling it not to recreate a specific existing video game. Original games, genre games, and games based on a user’s own designs and assets are intended to remain supported. For builders, the episode is a reminder that provider policy can alter an agent’s practical task boundary.

## Summary

Anthropic has confirmed that it added prompt guidance to Claude Code on Opus 5.5 telling it not to recreate a specific existing video game. Original games, genre games, and games based on a user’s own designs and assets are intended to remain supported. For builders, the episode is a reminder that provider policy can alter an agent’s practical task boundary.

## What happened

Game Developer reported on October 30, 2026 that Anthropic had updated Claude Code guidance to steer the product away from recreating someone else’s specific game. The reported rule says Claude should build original games, games in a genre, and games from the user’s own designs and assets. It should not build a recreation of a specific existing game, including one that is renamed or described as original.

The outlet initially attributed discovery of the rule to an X user who said it had appeared in Claude Code’s rules around October 27. Anthropic subsequently confirmed the change. According to the outlet, an Anthropic spokesperson referred it to a statement from Addy Osmani, a member of Anthropic’s technical staff. Osmani said the company had added two sentences to the Claude Code prompt on Opus 5.5 to steer the system away from recreating another party’s specific game.

Anthropic’s stated rationale is that using Claude to infringe intellectual property violates its usage policy. The company distinguished prohibited recreations from allowed work: original games, genre games, and games made with the user’s own designs and assets are meant to continue working. Osmani also said Anthropic would continue to iterate on the guidance to strike the right balance, and invited reports of cases where Claude declined work that belonged to the user.

The context is concern among game developers about rapid AI-assisted cloning. Game Developer says some studios have delayed public playtests because of fears that a demo could be ingested into an AI tool and used to create a competing clone. The report also notes developer pressure on Valve to explain how it will address clones on Steam. Anthropic’s intervention is therefore a product-level response to a specific misuse concern, rather than a new general-purpose game-development feature.

The reported mechanism matters as much as the subject. This is described as prompt guidance for Claude Code on Opus 5.5, not as a system that can independently verify copyright ownership or establish who owns an idea, design, or asset. The source itself raises that ambiguity. It also notes that prompt-level restrictions may deter low-effort copying without necessarily preventing determined users from seeking workarounds or choosing models with different restrictions.

## What the archive adds

We have not covered this specific policy change before; it opens a new thread.

## Trend signal

This event does not validate the claim that prompt guardrails are a durable enforcement boundary. It instead illustrates its limits: Anthropic describes a two-sentence prompt intervention, while the source questions whether restrictions can be bypassed and how ownership could be determined. That is consistent with the uncertainty identified in [Agent security moves to runtime boundaries](https://newruntime.com/patterns/agent-security-moves-to-runtime-boundaries/): prompts can shape behavior, but they do not by themselves establish reliable control over misuse.

## Why now

The source places the change amid recent anxiety about AI-assisted game clones. It says developers have worried that fast followers could reproduce an upcoming game from public material and reach the market first, and that some studios have delayed playtests in response. Anthropic’s confirmation came on October 30, after the rule was reportedly noticed around October 27.

## Why it matters

Teams using coding agents should treat a model’s apparent capability as conditional on provider policy as well as model quality. A workflow that successfully generated a kind of artifact yesterday can encounter a refusal after prompt or policy updates, particularly where requests resemble copyrighted works. That does not make the restriction unreasonable; it makes dependency visible.

The practical distinction is between legitimate work that happens to resemble a genre and a request framed as reproducing a named product. Teams building game tools, migration utilities, design-to-code systems, or creative generation features should make provenance explicit: retain evidence of user-owned assets, avoid describing owned work through another product’s identity, and provide an escalation path when a refusal is incorrect. The source does not show that such evidence will change Claude’s decision, but it is necessary operational context for human review.

## What happens next

- **Anthropic will refine the wording or handling of the game-recreation guidance after edge-case reports.** (3 months, 65%). Watch for: a documented clarification or changes to the stated allowed categories.
- **Builders of creative coding workflows will add more explicit ownership and provenance checks around agent requests.** (6 months, 55%). Watch for: product flows that collect asset rights or distinguish remakes from genre-based work.

## Limits and uncertainty

The source establishes Anthropic’s confirmation that two prompt sentences were added for Claude Code on Opus 5.5. It does not provide a complete technical specification of enforcement, a public changelog entry, error rates, examples across different prompts, or a way to test whether a user’s ownership evidence changes an outcome. It also does not establish how broadly the guidance affects Claude products beyond the stated Claude Code and Opus 5.5 context. Claims that prompt restrictions can be bypassed are reported as a general concern, not a measured evaluation of this particular rule.

## Skill to build now

This week, audit one agent workflow that creates derivative-looking output. Write down the source assets, the rights holder, the permitted transformations, and the reviewer who can resolve a mistaken refusal. Then test two task descriptions: one that names a third-party product and one that describes the required mechanics or visual constraints using only assets and requirements your team controls. The exercise will reveal whether your workflow is robust to policy-sensitive wording and whether it has sufficient provenance for a human escalation.

## Reader takeaway

Anthropic has drawn a prompt-level line in Claude Code between original or user-owned game work and recreating a specific existing game. The immediate rule is narrow, but the broader operational lesson applies to all coding-agent users: provider policy is part of the runtime environment, and workflows that rely on borderline tasks need provenance, review, and a fallback path.

## Open question

Can Claude Code reliably distinguish a user-owned remake or authorized port from a prohibited recreation when the relevant ownership evidence is outside the prompt?

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Claude](https://newruntime.com/topics/claude.md): Explore the claude topic hub.
- topic hub: [Anthropic](https://newruntime.com/topics/anthropic.md): Explore the anthropic topic hub.
- topic hub: [Claude Code](https://newruntime.com/topics/claude-code.md): Explore the claude-code topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)