🔍 Read the full analysis: What’s At Stake When AI Produces A Video Game Clone? on ThorstenMeyerAI.com
TL;DR
Take-Two CEO Strauss Zelnick said AI companies should protect others’ intellectual property, responding to a question about Anthropic’s restriction on recreating existing games in Claude Code. He called the issue a creative concern rather than a financial threat to Take-Two; the source material does not identify a specific clone or claim that either company enabled one.
IGN managing editor Wesley Yin-Poole asked Zelnick about claims that hackers were using AI to clone games faster and about Anthropic’s recent move to restrict game recreation in Claude Code. He also asked whether OpenAI should adopt a similar rule. Zelnick replied: “I think all companies should be protecting intellectual property, their own and everyone else’s.” He added that he considered it ironic for companies in the intellectual-property business to “stomp on someone else’s IP” while expecting their own work to be protected.
Zelnick did not specify what conduct he meant by “stomp on” another company’s intellectual property. He did not name a product, describe a particular clone or say that a Take-Two game had been recreated using Claude or ChatGPT. The interviewer named Anthropic and OpenAI, and Zelnick answered the question without making a specific allegation against either company.
Asked separately about AI-assisted piracy and the possibility of an unofficial PC version of Grand Theft Auto VI, Zelnick said he did not see it as a financial threat to the business but called it a creative concern. On Take-Two’s earnings call that day, he said piracy was likely to become “more of a challenge” and that the company was enhancing its technical defenses and legal activities. He did not provide an estimate of financial losses from AI cloning.
AI coding agents and existing games · as of Nov 14, 2026
Prompt in, game out: where the line runs
Three example requests (illustrative, not real cases) and what each vendor’s published rules say about them.
- Example request
„An original racing game with my own art“own idea, own assets
Claude Code (Anthropic)AllowedRule text: build original games and games from the user’s own designs and assets
Codex / policies (OpenAI)No objectionNothing in the published usage policies against it
- Example request
„A heist game with mechanics like [a named hit]“a game named as a reference
Claude Code (Anthropic)Allowed since v2.1.317Builds an original game in that genre, per Anthropic
Codex / policies (OpenAI)No game rule foundNo game-specific instruction in the Codex CLI code
- Example request
„Rebuild [a specific game], just rename it“a copy under a new name
Claude Code (Anthropic)Not allowedCovers its name, characters, art, music, levels or look; offers to build something similar
Codex / policies (OpenAI)General IP banUsage policies bar attempts to infringe IP; no game-specific rule found
Outside the tools: Take-Two says it relies on its own „technical defenses“ and „legal activities“. Copyright and trademark law apply whichever tool writes the code.
The requests are our own illustrations. The rows summarize published rules, not tests of how the tools behave. ChatGPT’s and Codex Web’s server-side instructions are not public.
Sources: Claude Code program text, versions 2.1.317 to 2.1.325, and Anthropic’s statement of Nov 4, 2026 (via Tom’s Hardware); OpenAI usage policies, version effective Nov 20, 2026 (Internet Archive copy of Oct 28, 2026); Codex CLI repository on GitHub, checked Nov 14, 2026; Take-Two fiscal Q2 2027 earnings call, Nov 9, 2026.
Graphic: Powered by Thorsten Meyer AI
The Stakes for Game Creators
The discussion concerned whether AI models could help produce close copies of existing games and whether AI companies’ rules address intellectual-property rights. Zelnick said companies should protect those rights but did not establish that any specific company or tool had infringed Take-Two’s rights.
Potential copies could be distributed outside a publisher’s intended release and service arrangements. For developers and publishers, the issues include control over characters, art, music, levels and other elements of a game. Zelnick described the concern as creative rather than financial, saying he did not view it as a financial threat to Take-Two at that time.
A policy restricting a model’s output does not establish that the restriction prevents all copying. The existence of a policy also does not, by itself, resolve questions about how a model was trained or how users apply it. The interview raised these issues but did not provide evidence to resolve them.
Popular right now

Bose QuietComfort Ultra Earbuds (2nd Gen), Wireless Noise-Cancelling Earbuds, Black

De'Longhi Magnifica Start Automatic Espresso Machine with Milk Frother, Silver

Sony WH-1000XM6 Wireless Noise-Canceling Headphones, Black
As an affiliate, we earn on qualifying purchases.
What the AI Rules Actually Say
Anthropic added a game-specific instruction to Claude Code in version 2.1.317, published Nov. 2, according to the source material. It tells the model not to create a recreation of a specific existing game using its name, characters, art, music, levels or look, including if the output is renamed or described as original. Anthropic grounds the instruction in its usage policy, which prohibits infringing third-party intellectual property.
OpenAI’s usage policies also prohibit attempts to infringe others’ intellectual-property rights. The source material says its author found no game-specific instruction in the open-source code of OpenAI’s Codex CLI. That observation is limited to the code reviewed; it does not establish what other safeguards OpenAI may use across its products.
The interview was published the evening Take-Two reported results for the quarter ended Sept. 30, 2026. On the earnings call, Zelnick said technical limitations had historically made it difficult to replicate the experience of playing Take-Two’s games as intended, citing factors such as file sizes, piracy complexity and server connections. He also said piracy could become more challenging as technology changes.
„“I think all companies should be protecting intellectual property, their own and everyone else’s.”“
— Strauss Zelnick, Take-Two Interactive CEO, speaking to IGN
What the Interview Does Not Establish
The interview does not establish that Claude Code, ChatGPT or another named AI tool has been used to clone a Take-Two game. It also does not identify what Zelnick meant by “stomp on someone else’s IP”: he did not clarify whether he was referring to material generated by users, model development or another practice.
The interviewer’s reference to reverse engineering is distinct from Anthropic’s published rule. As described in the source material, the rule addresses recreating a specific existing game; it does not explicitly mention decompiling or reverse engineering. The source material reports no public response from Anthropic or OpenAI to Zelnick’s remarks as of 21:30 UTC on Nov. 14. The scope and effectiveness of the companies’ safeguards, and any particular alleged infringement, remain unestablished here.
Safeguards and Piracy Defenses
Take-Two said it was strengthening its technical defenses and legal activities, but the source material does not give a timetable, describe specific measures or report a new enforcement action. The company’s next steps, and whether its approach changes as AI tools develop, remain to be seen.
For Anthropic and OpenAI, the open question is whether either company will publicly address Zelnick’s comments or explain how its rules apply to game recreation and other potential forms of copying. Any later assessment should distinguish a stated policy from evidence that a specific tool produced an infringing game or that a rights holder pursued a case.
Key Questions
Did Zelnick accuse Anthropic or OpenAI of cloning a Take-Two game?
No specific accusation appears in the interview as described in the source material. Zelnick made a general point about protecting intellectual property after the interviewer named both companies.
What does Claude Code’s game rule prohibit?
Anthropic’s instruction tells Claude Code not to recreate a specific existing game using its name, characters, art, music, levels or look, including under a different name. The source material says it was added in version 2.1.317, published Nov. 2, 2026.
Did Zelnick say AI cloning threatens Take-Two’s finances?
No. He said he did not view it as a financial threat to the business, while describing it as a creative concern. He separately said piracy could become more challenging.
Has OpenAI announced a game-specific restriction?
The source material says OpenAI’s usage policies bar attempts to infringe others’ intellectual-property rights, but its author found no game-specific instruction in the open-source Codex CLI code. That finding does not establish what safeguards may exist in other OpenAI products.
Source: ThorstenMeyerAI.com