Anthropic stops Claude Code from cloning other companies' games: a legal precaution with an IPO in sight

By GamesBeat (with coverage from Game Developer and Engadget) · November 1, 2026.

Anthropic has modified the system prompt of Claude Code, its coding agent, so that the model does not recreate someone else's specific game. This was confirmed by Addy Osmani, a member of the company's technical staff, in a post on X on October 30. According to Osmani, it amounts to “two sentences” added to the prompt on Opus 5.5, and the reason is that using Claude to infringe intellectual property violates the company's usage policy.

The change came to light when an X user identified as Chasm spotted an update to Claude Code's rules instructing the model not to build a replica of an existing game, even if it is renamed. Game Developer reported on it and the company pointed to Osmani's post.

Osmani clarified the scope: original games, genre games and those made with their own designs and assets should keep working. He added that they will continue iterating to “find the right balance.” A system prompt is the set of hidden instructions that guides the model's behavior. This is therefore guidance, not a technical block.

The context, according to GamesBeat, is a wave of AI-made clones in recent months. That wave has led some companies to give up public playtests of their games. Asking Claude Code for a “match-3 like Candy Crush” quickly produces something functional, but that does not shield whoever publishes it from legal consequences.

The outlet also notes that Anthropic may be reducing its exposure to copyright infringement claims from the video game industry. This is the outlet's interpretation, not a statement from the company. It coincides with the timing: according to GamesBeat, Anthropic filed its S-1 form (the public registration to go public) on October 26. That document would reportedly show revenue of about $16.3 billion in the first half of 2026, 14 times more than a year earlier, and a net loss of $80.7 billion (with an adjusted operating loss of $693 million).

**Our take.** My thesis is that this change is less about video games than a symptom of maturity. When a tool goes from curiosity to production infrastructure, its makers begin to take responsibility for what is built with it. Ahead of an IPO, every litigation risk weighs on valuations.

There is also an underlying tension. The line between “cloning a specific game” and “making a game in the same genre” is blurry: mechanics are generally not protected by copyright, whereas code, art and assets are. A two-sentence prompt cannot draw that boundary precisely, and there are likely to be false positives and ways around it. The company's own talk of iterating points in that direction.

The winners are studios, which see a legitimate concern recognized, and Anthropic itself, which reduces its risk. The losers are those who used the agent to prototype “in the style of” a reference title, even for legitimate purposes. In the short term, these frictions are the price of integrating generative AI into creative industries with owners and rules.

In the long term, the lesson is positive. An AI capable of producing complete software in minutes makes creation cheaper, and that points to more people making original games with their own designs, which is exactly what Osmani says he wants to preserve. Negotiating intellectual property limits early and judiciously is healthier than leaving them to the courts later.