As generative AI becomes increasingly commonplace in game development, Valve created a Steam tag to label any releases that use such tech. In a new wrinkle, this labelling now pertains more explicitly to public assets, distinguishing them from other uses of the tools.
Simon Carless of GameDiscoverCo noticed the discrepancy in Steam's backend. There have been some alterations to the questions asked. "Does this game use generative artificial intelligence to generate content for the game, either pre-rendered or live-rendered?" read the original text. "This includes the game itself, the storepage, and any Steam community assets or marketing materials."
This was followed by checkboxes specifically for pre-rendered and live-rendered, and a text box to provide further details. All cut and dry and understandable. The questioning has since been altered for wiggle room in terms of any given studio's pipeline.
"Efficiency gains through the use of these tools is not the focus of this section," the page currently reads. "Instead, it is concerned with the use of AI in creating content that is shipped with your game, and is consumed by players."
From then on, the original phrasing is largely intact. It seems that Valve is attempting to focus on what tags help consumers understand what they're buying, while making allowances for developers who are using AI in places away from players.
There's been discussion around utilizing bots to assist in the concept phase. Baldur's Gate 3 developer Larian Studios mentioned doing so for visuals and flavor text, though the latter seems to have since been thrown out for being fundamentally sub-standard.
Valve's introduction of AI tags was itself controversial, prompting comments from Epic's Tim Sweeney and others about what's ultimately useful to users. I doubt the custodians of PC's biggest marketplace have seen the end of the AI conversation, nor will this be the last modification made to its criteria around such in-game development.
