Sony Plans to Use AI to Catch Predators and Scammers on PlayStation Network

Sony Interactive Entertainment has filed a patent for a safety system designed to detect malicious users in online games and digital services using artificial intelligence honeypots. The patent application, numbered 18/314775, was published on August 13, 2026.

At the center of the technology is an AI-powered decoy. The system deploys a pseudo-account designed to mimic a real user, matching specific demographics and conversational styles. The patent specifically outlines scenarios in which this decoy poses as an underage player.

When a suspicious user begins interacting with the fake account, the system analyzes the conversation and user behavior. The patent notes that the decoy account can be driven by a large language model, such as GPT.

Sony describes the framework as functioning similarly to a police sting operation adapted for virtual spaces. Merely engaging with an AI account would not result in an automatic ban. Instead, the interaction serves as an indicator of suspicious intent.

The patent outlines several high-risk behaviors the system is designed to identify:

• Attempts to move conversations to private channels
• Efforts to extract personal information
• Unsolicited contact directed at users appearing to be minors
• Sending inappropriate messages
• Attempts to scam users or steal personal data
• Spamming activity
• Behavior linked to cheating and exploits

If the algorithm flags suspicious activity, the platform can place the user account under closer monitoring. When that flagged user later attempts to interact with actual players, the system can issue proactive warnings or apply protective restrictions.

Current moderation systems usually depend on user-submitted reports, meaning a bad actor often has to contact a minor or victim before an investigation begins. Sony’s proposed approach aims to spot and isolate threatening behavior before a bad actor reaches a real player.

While the patent places major emphasis on child safety, the underlying technology has broader applications. The document notes that the honeypot mechanism can also target fraudsters, spammers, cheaters, and other users exhibiting disruptive online behavior.