PlayStation Patents AI “Pseudo Account” to Flag Potentially Hazardous Users

Sony Interactive Entertainment's PlayStation division has filed a patent, numbered 18/314775, outlining a “Large Language Model Powered Social Integrity System.” The system introduces a pseudo account designed specifically to identify and flag potentially dangerous individuals on its platforms without the suspect knowing they are interacting with an automated moderation tool.

According to the patent documentation, the pseudo account is configured to mimic the demographic messaging behavior of a potential victim—such as a minor—making it appear completely legitimate to anti-social users. Once an anti-social account establishes communication with the pseudo account, the system flags the account as suspicious. It then waits for the anti-social user to initiate a move on a legitimate account and report it before undesirable communication can occur. The pseudo account itself can be controlled by a large language model, such as a generative pre-trained transformer.

The system is designed to detect various forms of suspicious activity, including spam, requests for personal information, inappropriate messaging, in-game cheating, scams, and attempts to contact potential minors. Rather than immediately banning or punishing the user, the system flags those who meet the conditions so that anyone who subsequently interacts with them will see a warning. However, as with all patent applications, it remains unclear whether this technology will eventually materialize in a commercial product.

In related corporate news, PlayStation CFO Lin Tao recently addressed the company's decision to stop producing physical discs for new titles starting in January 2028. Acknowledging significant backlash regarding the upcoming hardware transition, Tao stated that the company would cautiously move the initiative forward.