Anthropic restricts Claude from recreating existing games amid AI cloning concerns / Juicebox

19
Anthropic restricts Claude from recreating existing games amid AI cloning concerns (pocketgamer.biz)(pocketgamer.biz)
byindustrybot4 days ago4d ago

More from pocketgamer.biz

Also covered by

Anthropic orders AI tool Claude to stop remaking video games (gamedeveloper.com)
Anthropic adds rule to stop Claude Code recreating existing games (gamesindustry.biz)
Claude's new rules against game cloning won't stop the AI copying boom (arstechnica.com)
Fewer genAI clones of "someone else's specific game" should flood the net going forwards, as Anthropic tell Claude to stop messing with some stuff other people own (rockpapershotgun.com)
Anthropic's Claude AI blocks video game cloning and reverse-engineering (tomshardware.com)
Anthropic updates Claude’s coding instructions to ban 1:1 game remakes like Minecraft and GTA due to AI slop and IP concerns, so no reverse-engineering, but developers can still use it to build original games from scratch. (reddit.com)
Anthropic takes steps to block Claude from reverse-engineering video games (mashable.com)
Anthropic may be cracking down on use of Claude to remake popular games (gamesbeat.com)
Anthropic's Claude blocks video game cloning and reverse-engineering (mastodon.online)
Anthropic Builds World's Tiniest Dam Against AI Clone Game Flood (aftermath.site)
Anthropic Cracking Down on Video Game Rewrites (twitter.com)

Sign in to comment.

0

tldrbot3 days ago

TL;DR of the article, generated automatically.

Anthropic has confirmed it tweaked Claude Code to steer it away from building copies of existing games, after a user spotted the change.

What changed: Two sentences added to Claude Code's prompt on Opus 5.5 steer it away from "recreating someone else's specific game."
What's allowed: Original games, genre games, and games built from your own designs and assets, per Anthropic's Addy Osmani.
Timing: The rule was reportedly pushed October 27, and Chasm reports there is no way to disable it.
Doubts: The article says safety limits can be sidestepped with rhetorical tricks, ownership is hard for the software to judge, and open-weight models have no such rules.

The article's bottom line: this may only stop the lazy copycats.