Following the recent ban on titles featuring AI-generated artwork, Valve has made a move to clarify its stance on games with AI-generated content.
Last week, a developer asked to certify that a game featuring AI-generated assets as an initial template owned all necessary rights to be used in the datasets it created by training the AI. Claimed it was rejected by Valve because he requested it.
More and more developers are using AI to create games. But the rapid proliferation of generative AI tools trained on man-made art scraped from the web has raised copyright issues that didn’t exist before.
In the wake of last week’s incident, Valve released a statement to VGC, stating that “its purpose is to: [games built with AI] On Steam,” he said, adding that the actions taken so far “reflect current copyright laws and policies and do not add to our opinion.”
“We continue to learn about AI, how it’s used in game development, and how it can be incorporated into the process of reviewing games submitted for distribution on Steam,” it reads. “Our priority, as always, is to try to ship as many of the titles we receive as possible.
“The introduction of AI can make it difficult for developers to demonstrate that they have sufficient rights to use AI to create assets such as images, text and music.”
A spokesperson pointed to “legal uncertainties regarding data used to train AI models” and said it was the developer’s responsibility to “have the proper rights to ship the game.”
“We know this is an ever-evolving technology, and our goal is not to prevent it from being used on Steam. Instead, we’re looking at ways to integrate it into our existing review policy. Therefore, our review process reflects current copyright laws and policies and does not add to our opinions.As these laws and policies evolve over time, our The process will evolve.
“We welcome and encourage innovation, and AI technologies are sure to create new and exciting experiences in games. , cannot infringe any existing copyright.”
AI creation is still a gray area in many parts of the world, with governments, artists, and companies weighing how best to move forward, but Japan has recently declined to allow the use of datasets to train AI models. Declared not to violate copyright law.
As reported by Decrypt, this decision means that model trainers can use publicly available data without securing permission from the data owner.
