Anthropic's Claude AI blocks video game cloning and reverse-engineering

Discussion overview

The thread quickly moved past whether Claude should clone games and into who gets to control capable coding models. A minority argued that cloning is creatively hollow: it lets an LLM replay decisions already made instead of helping developers explore untried mechanics. Even that camp saw modding as a productive bridge, where a familiar game becomes a scaffold for learning and eventually for genuinely new work.

Most of the energy ran the other way. Commenters treated reverse engineering and ports as legitimate preservation tools, especially for abandoned online games, obsolete platforms, and versions altered beyond recognition. They argued that cultural artifacts and communities disappear when rights holders shut servers or stop supporting old software, and AI could make resurrection and modernization practical for ordinary users. Several also saw the restriction as a legally defensive but intellectually inconsistent policy: hosted vendors can forbid explicit reverse engineering while still assisting adjacent forms of imitation.

The deeper fear was dependence on a small number of hosted-model gatekeepers. Readers expect people needing these capabilities to migrate to competing or local open-weight models rather than abandon the work. Some warned that unfiltered local models bring their own security risks, but the prevailing view was that prompt-level controls will not hold once capable models are widely runnable. The practical alternative proposed was better provenance and reputation systems, not trying to make copying impossible.

Points of disagreement

The central fight is whether blocking game cloning protects legitimate rights or prevents legitimate preservation and user autonomy. Supporters of the limit see automated copying as an easy substitute for design work and a natural accelerator for infringement; opponents argue that old builds, private servers, ports, and mods preserve software culture that publishers routinely abandon. The latter side says vendor restrictions merely redirect users to less constrained models while concentrating power in a few AI providers.