by 
Anthropic, the owner of Claude Code, quietly added new rulings to the AI bot's programming this past weekend, seemingly in an effort to prevent its widespread use in the cloning, reverse-engineering, and decompilation of existing games.
As spotted by various users on Twitter/X, the newly established Line 2397, denoted "tengu_heron_brook", requests that Claude "(Does) 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." The new line of code cannot be removed or bypassed by users.
While Anthropic did not release a statement on the change, the reasoning seems relatively clear. Over the past year in particular, Claude has been used in thousands of homebrew projects that focus on the decompilation and recompilation of published software - either for basic modding purposes, translation and asset adjustment, or simply outright copying and reselling of the same game, albeit with a few minor titling adjustments.
Last week's Steam Next Fest brought the problem to mainstream attention, with the usually joyous demo festival soured by the addition of thousands upon thousands of low-quality, half-arsed, and wholly derivative games. Some were clones of existing titles, and others were clones of clones. The impact was felt by players, developers, and Steam owner Valve itself, with many critics, devs, and players dubbing the event a variant of the "worst" or most "depressing" Steam Next Fest yet.
Now, to be clear, it is my opinion that Anthropic, like most AI bot developers, give zero shits about users infringing on the copyright of others, and so the sudden implementation of the line suggests (and it IS just speculation) that someone, internally or externally, has leaned on the company to take steps (or be seen to be taking steps) against users employing Claude in the cloning of video games and the facilitation of code theft.
Of course, this move is closing the door when the horse has already bolted. While Anthropic's action will put up an extra barrier to anyone who falters at the first hurdle (many bad-faith AI users in a nutshell), there are already a plethora of other AI tools out there that feature no such restriction. While a quick look on Twitter/X reveals many Claude users who are furious at Anthropic's decision, others are declaring their intention to simply move to another tool or just develop a workaround.
The reality is that gen-AI is built upon the concept of plagiarism. Whether asking it to create Sonic vs. Simon Belmont vs. John Wick Racing or simply asking for tomorrow's weather, the result comes from the consent-free scraping and repackaging of others' work. Anthropic's decision has likely only been made in an attempt to cover its own ass, lest somebody, or Some Body, eventually comes crashing down on the entire AI industry.
Ultimately, Line 2397 is barely a restriction. And tantamount to trying to prevent a tornado by putting up a windbreak. Without a legitimate form of industrial regulation, the slop conveyor belt will churn on. And, frankly, I don't see that regulation rolling over any kind of horizon.