What leaked “SteamGPT” files could mean for the PC gaming platform’s use of AI

These days, it seems like every tech company and their corporate parents are looking to incorporate AI tools and features into their products, whether they want to or not. So when files with names and functions referencing “SteamGPT” appeared in a recent Steam client update, Valve watchers took quick notice.

From the outside, it’s hard to tell exactly what form such a “SteamGPT” would take. But looking at the variable names and references in the files suggests that Valve is looking to use AI tools to streamline internal assessment of in-game incidents and investigate potentially suspicious accounts.

looking at variables

As tracked by the Automated SteamTracking GitHub project, the term “steamgpt” appears multiple times in three different files added in the April 7 Steam client update. In addition to the SteamGPT naming convention – an apparent reference to the generative pre-trained transformers popularized by ChatGPT and its ilk – the files include mentions of terms like multi-class estimation, fine-tuning, and “upstream models” that point to some type of generative AI system.

What can that AI be used for? Well, the files contain a labeler and several references to “labeling tasks”, which work with arguments identifying a “problem” and a “subproblem” and looking at the “evaluate_evidence_log” related to a specific “matchid”. With the mention of the “logs_to_inference” metamodel, it seems like this may be a hook into the system to automatically generate labels to classify various incident reports made in Steam multiplayer games.

steamgpt

The SteamGPT function references several security features.

The SteamGPT function references several security features.


Credit: Steamtracking/Github

Another part of the SteamGPT files hints that Valve may be using AI tools to summarize suspicious activity history or patterns in potentially fraudulent Steam accounts. Several “SteamGPTSummary” functions include references to VAC bans, Steam Guard, and account lockdowns. These functions also look at evidence such as email address (“high_fraud_email”), use of advanced security features (“two_factor”), and where a linked phone number originates (“phone_country”) to help determine whether an account is on the up-and-up. There are also some references to account trust scores, which are already used in games to help secure matchmaking counter attack 2.



<a href

Leave a Comment