So far there’s only been hints that game-maker and Steam-owner Valve has been dabbling in generative AI use. But if recent findings are to be believed, it might be working on a “SteamGPT” AI feature to help tackle cheating on Steam and manage account restrictions.
Valve-focused dataminer Gabe Follower (via Tom’s Hardware), who traditionally works through Steam code updates looking for new information, unearthed code recently that mentioned “SteamGPT” and did so in relation to account statistics and Steam’s Trust Score metric. This helps Valve deal with cheating and account restrictions in online games connected to Steam.
As pointed out by Gabe Follower, the code appears to be tied to several account stats such as age, confidence score, model evaluation, and even pre-existing bans as variables. If Valve is indeed working on an AI helper, it seems the goal here is to improve the efficiency of the process of dealing with customer support tickers and recurring issues like false bans.
Valve has struggled with cheaters on Steam before, and its technology hasn’t always been up to the task. In late 2023, Valve’s Anti-Cheat (VAC) System doled out false bans for simply moving your mouse too quickly in Counter-Strike 2. Meanwhile, cheating continues to be an issue that Valve is battling against, so AI automation could be appealing, even if the divisive tech continues to raise concerns.
Nevertheless, this unearthed code could only be something Valve is testing but won’t release. And even if it does make it through to release, “SteamGPT” might remain an internal tool rather than anything consumer-facing.