Valve Clarifies Policy on AI-Generated Assets
Valve has released a rare statement in response to allegations that it rejects games featuring AI-generated assets from its Steam games store. The company clarified that its policy is evolving and not a direct opposition to the use of AI. Like the other app platforms, Steam follows a review and approval process with content guidelines that are not always explicit. However, one indie developer claimed that Valve was no longer publishing games with AI-generated content.
It was unclear who held the rights to the AI-trained data used in their creation. As a result, Valve got concerned regarding the legal ownership of such assets. The developer mentioned that the game they had submitted contained a few AI-created assets which further raised concerns for Valve.
In the warning letter sent to the developer, Valve stated, “As the legal ownership of such AI-generated art is unclear, we cannot ship your game while it contains these AI-generated assets unless you can confirm that you own the rights to all the intellectual property used in the dataset that trained the AI.”
A week later, Valve reviewed the game further and informed the developer that games can not be distributed if the necessary rights are not obtained. In this case, Valve declined to distribute the developer’s game due to uncertainty regarding the rights to the training data of the underlying AI technology used to create the assets.
This policy essentially establishes a broad prohibition on AI-generated assets in games. Most AI tools cannot definitively claim legal rights to all their training data. Even if they could, there are ethical concerns surrounding the use of such data.
While the use of AI as a game development tool is not inherently controversial, there is a distinction between AI assisting in creating realistic environments or altering quest structures and generative AI driven by unpaid artists. The potential legal liability for creators, distributors, and other parties involved in AI-generated art remains unclear. Valve avoids publishing AI-generated content if creators cannot establish clear copyright ownership, deeming the associated risk too high.
In response to Eurogamer, Valve clarified that its policy focuses more on adhering to legal requirements rather than taking a particular stance on AI. The statement read,
“We recognize AI as an ever-evolving technology, and our goal is not to discourage its use on Steam. Instead, we are working on integrating AI into our existing review policies. Our review process reflects current copyright laws and policies, rather than imposing an additional layer of personal opinion. As these laws and policies develop, so will our process.”
Meanwhile, Valve has announced that it will refund the app submission fee, typically non-refundable, in cases where this evolving policy becomes the determining factor.
At present, it remains unclear whether AI is being used extensively outside of experimental contexts or, as in the case mentioned earlier, as a means for blatant profiteering. The developer in question posted a video purportedly showcasing their development process, exclaiming, “Get rich quick with AI waifus” and “How I made $1000 Publishing an AI-Generated Game on Steam (Gone Sexual?!?!)”. Consequently, the impact of Valve’s policy may not be heavily felt. However, as more established developers adopt these tools and their sophistication increases, the situation may become less straightforward and more complex.