According to BeInCrypto, Anthropic added two sentences to the code underlying its Claude model to prevent the generation of content that replicates video games. The update aims to stop AI systems from producing outputs that closely resemble existing video game elements such as storylines, mechanics, or assets.
Developers have reported that the change also interferes with legitimate internal development work, including when teams use AI tools to explore concepts for their own games. Some developers say the filter activates during routine brainstorming or prototyping, even when no copying of third-party games is intended.
Anthropic did not provide technical details about the wording of the two sentences or how the filter determines what constitutes a clone. The company also has not released an updated policy document outlining the change.
