Leaked Steam files suggest Valve is testing an AI security review tool

In short: Files found in a recent Steam update suggest Valve may be testing an AI tool to help staff review player reports and suspicious accounts.

What happened

People who track changes in the Steam app noticed new files in an April 7 Steam client update that repeatedly mention “SteamGPT.” These files were spotted through the SteamTracking project on GitHub, which automatically logs what changes in Steam updates.

The files include names and references that point to a text generating AI system, similar in concept to tools like ChatGPT. The code mentions things like “labeling tasks” and “evidence logs” tied to specific game matches. In plain terms, it looks like Valve may be trying to use AI as a sorting assistant, like a mailroom worker who reads envelopes and puts them into the right bins.

Other parts of the files suggest a second use. They describe AI summaries connected to account security signals such as VAC bans (Valve’s anti cheating bans), Steam Guard (an extra login check), account lockdowns, two factor login, and where a phone number is from. The goal appears to be helping staff quickly review whether an account might be fraudulent.

Valve has not confirmed that “SteamGPT” is active, or that it will ship to users. Ars Technica notes that these could be prototypes that never get used.

Why it matters

Steam has a huge number of players, reports, and account disputes every day. If Valve uses AI to help review this information, it could speed up decisions and reduce manual work. It also raises questions players care about, like how accurate those automated summaries are and how often real people double check them.

Source: Arstechnica