Sony Patent Filing Describes AI That Adapts NPC Behavior To Player Style
The filing sketches a combat and conversation system driven by per-player telemetry, and the only public reaction so far is whether adaptive difficulty is something players would want or should be able to switch off.
Reporting from 1 source: Game Spark.
A Sony Interactive Entertainment patent application filed March 31, 2025 and published October 1, 2026 describes using AI to change NPC and boss behavior based on player actions. It lists weapons used, dodge and parry success rates, attack timing, and retry counts as collected data, fed into a machine learning model that generates NPC strategies and responses. The document names no specific title or timing.
The application, titled "AI-ADAPTED NON-PLAYER CHARACTER FOR VIDEO GAME SYSTEM," describes encounters between the player and NPCs being collected from the console, then fed into a machine learning model that produces NPC behavior.
The listed inputs go beyond wins and losses: weapons and skills used, attack frequency and timing, positioning and dodging patterns, and the frequency and success rate of dodges, guards, and parries. The document extends past combat to NPCs that flee, fight, or negotiate based on the player's aggressiveness or in-game reputation, and to conversation options and NPC roles that shift. No title or implementation window is named.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.