[Digital Today Kyoung-min Hong (홍경민), intern reporter] Valve, which operates the global PC game platform Steam, appears to be preparing to introduce a generative AI technology called SteamGPT to improve internal operating efficiency, including in-game incident reports and identification of abusive accounts.

On April 10 (local time), IT outlet Ars Technica reported that numerous pieces of reference data linked to a new feature named SteamGPT were spotted in a recent Steam client update file.

Files confirmed through GitHub's Steam Tracking project were found to include many terms closely tied to generative AI systems, such as multi-category inference, fine-tuning and upstream models. This is seen as a move by Valve to fully apply pre-trained transformer technology, which came to prominence with ChatGPT, to the operation of its platform.

Analysis of the leaked files suggests SteamGPT's main role will focus on streamlining system operations and monitoring rather than an interface directly exposed to users. In particular, many references were found related to a labeler function that analyzes and classifies in-game incident reports. Based on evaluation evidence logs linked to a specific match identification number (Match ID), it automatically categorises where a problem occurred. This points to an intent to quickly process a huge volume of user reports generated in multiplayer games.

It was also confirmed that a SteamGPT Summary function exists that summarises the activity history of suspicious accounts. The function reviews various security-related variables, including Valve Anti-Cheat (VAC) ban history, whether Steam Guard is enabled and account lock status. It is likely to be used as an auxiliary tool to judge account trustworthiness by comprehensively analysing whether a high-fraud-risk email address is used, the extent of two-factor authentication use and nationality information tied to a linked phone number. This is expected to further advance the trust score system used for matchmaking security in games such as Counter-Strike 2.

Valve's move aligns with management's philosophy that AI technology will have a significant impact across business operations. Gabe Newell (게이브 뉴엘), Valve's chief executive, has previously compared the growth of AI to the advent of spreadsheets or the internet and stressed that machine-learning systems will deeply penetrate every area of business. He has also said developers can create higher value when using AI as a tool. Valve has in practice shown an active stance on adopting the technology, explicitly allowing the use of AI tools in the development process for games released on Steam since 2024.

Still, it remains unclear when and how the files found this time will be applied to an actual service. For now, it appears less like a gamer-facing tool such as Microsoft's Gaming Copilot and more like an internal back-end tool to help platform administrators effectively filter through piles of incident logs and abusive accounts.

As AI influence expands across the ecosystem, including with about 8,000 Steam releases already being developed using AI technology as of mid-2025, Valve's experiment in introducing AI is seen as an important turning point for ensuring service integrity.

It seems that Valve is working on a "SteamGPT" feature that will apparently deal with Steam support issues and is somehow connected to Trust Score and CS2 anti-cheat? pic.twitter.com/a3MckicQf2