This period saw 41 AI and machine learning patent applications filed across 14 companies, led by Sony (19), Nvidia (5), Tencent (4), Microsoft (3), and single filings from Bandai Namco, EA, Roblox, AMD, Acco Brands, Adeia, Atombeam Technologies, Wizards of the Coast, AT&T, and Triumph Labs.
Sony's applications focus heavily on LLM-powered in-game assistants that provide personalized guidance through character voices, AI systems that customize audio and haptic feedback based on player behavior, and tools for generating 3D assets and personalized content. Nvidia filed patents covering conversational AI for NPCs, including systems that insert images into chatbot responses and process spatial audio for realistic 3D interactions, while Microsoft described AI assistants that detect player struggles and provide contextual help based on expert gameplay data. Tencent and AMD addressed NPC behavior through reinforcement learning systems, EA covered AI-driven task assignment for multi-character coordination, and Roblox detailed machine learning approaches to cheat detection in user-generated content platforms.
Sony received 19 patents covering a wide range of AI-powered gaming experiences. Several address in-game assistance through LLM-powered systems that let players ask video game characters for story recaps, gameplay tips, and navigation guidance delivered in the character's voice, with one patent describing an AI teammate that monitors health, tracks enemies, and suggests strategies in multiplayer games, and another enabling similar support for solo players through real-time strategic advice. The company filed multiple patents around personalization, including an audio system that learns player preferences to automatically modify or replace in-game sounds, a trophy customization system that adapts achievement appearances based on player behavior and playstyle, and an NPC dialogue system that analyzes how players interact with characters to dynamically adjust responses across multiple sessions. Sony also described systems for content generation, including one that creates personalized 3D game items in under 2 minutes using player data and neural rendering, another that auto-generates transitional video between gameplay moments for seamless highlight reels, and a third that analyzes gaming video to create 3D overlays highlighting popular paths based on aggregated player behavior. Additional patents cover accessibility features like automated narration generation that inserts audio descriptions during dialogue gaps with emotion matching, haptic feedback that provides directional cues to struggling players while tracking achievements separately, and a pitch control algorithm for fine-tuning text-to-speech voices to enable dynamic character voice customization. The company filed 2 patents for AI-generated podcasts that deliver game recommendations and gaming news in the voices of players' favorite video game characters, plus patents covering automated esports broadcasting with dynamic camera angles and real-time narration, a proactive assistant that predicts player questions by analyzing gameplay video frames, an eye-tracking system that suggests chat initiation between multiplayer players based on gaze focus, a player engagement monitoring system that signals through wearable lights whether someone is receptive to interruptions, and a haptic rendering system that conveys detailed physical properties like material composition and surface characteristics for metaverse interactions.
Nvidia received 5 patents spanning conversational AI and content generation. One enables chatbots and NPCs to insert relevant images into text responses by pre-indexing text-to-image associations and using vector similarity search, avoiding the computational cost of multimodal models. Another processes multichannel spatial audio by converting arbitrary audio channels to a fixed format, allowing AI agents to perceive and respond to directional sound cues in 3D environments. The company filed a patent for an auto-regressive auto-encoder that generates artist-quality 3D meshes with up to 8000+ faces through efficient face tokenization and compression into fixed-length latent codes. Nvidia also described an NLP system that extracts game events from text logs rather than video frames for real-time highlight generation, plus a real-time censorship system using deep neural networks to filter abusive language in multiplayer games and live streams with low enough latency to work during active communication.
Microsoft filed 3 patents focused on AI-powered player assistance. One trains machine learning models on expert gameplay to create adaptive AI assistants that can actively play through difficult sections in real-time, moving beyond static video tutorials. Another automatically detects when players struggle by analyzing aggregate gameplay data and offers proactive contextual help rather than waiting for manual help requests. The third intelligently restricts what actions a helper can perform during coaching sessions based on game state, preventing spoilers while still allowing guidance through filtered input controls.
Tencent received 4 patents addressing customization and NPC behavior. One lets players generate custom clothing designs in games using keywords and original garment templates, maintaining technical compatibility with game engines while offering unlimited options beyond presets. The company filed 2 patents on NPC control, including a system that separates individual and group behavior layers to enable complex, dynamic AI responses, and another that uses reinforcement learning with LSTM networks to train NPCs for realistic combat decisions, replacing manual behavior tree programming. Tencent also described an automated verification system that tests user-generated game levels for playability using AI-driven pathfinding, identifying navigation errors and offering one-click fixes that adjust level geometry.
Nvidia's patent on AI-driven task assignment enables dynamic optimization of character-to-location matching in multi-character video games. EA filed a single patent describing a ring-based assignment algorithm that scales efficiently when multiple characters must be matched to multiple targets, solving the coordination problem holistically across all characters and objectives simultaneously rather than handling each independently.
Roblox filed 1 patent for a machine learning anti-cheat system that detects player cheating by analyzing game state data and physics violations in real-time. The system uses game-specific or genre-specific models that can detect multiple cheat types through a single model or ensemble, moving detection from unreliable client-side monitoring to server-side ML inference with integration to language models for automatic generation of cheat-resistant scripts.
AMD received 1 patent for a hierarchical reinforcement learning framework where leader NPCs train follower NPCs through demonstrations. This approach reduces computational resources by having experienced NPCs act as teachers rather than training every character from scratch, while maintaining diverse and adaptive behaviors across different skill levels and roles.
Acco Brands filed 1 patent for an AI-powered game controller with a built-in virtual assistant that provides real-time gameplay tips and strategy guidance. The system uses computer vision to recognize games via camera, listens to gameplay audio, interprets natural language commands, and can execute complex in-game actions through voice control.
Adeia received 1 patent for an NPC personality system that uses relationship graphs and LLMs to make every character in a game world dynamically respond to player actions. The system scales interactive NPCs beyond manually scripted characters through graph-based personality propagation while preventing information leakage through knowledge base constraints tied to player progression and NPC relationship strengths.
Atombeam Technologies filed 1 patent for a Persistent Cognitive Machine that creates AI with continuous thought processes and memory that persists across system restarts. The system introduces biologically-inspired sleep cycles for memory consolidation and autonomous thought generation independent of external prompts, enabling synthetic teammates that evolve with gaming communities rather than operating on a prompt-response paradigm.
Wizards of the Coast received 1 patent for an AI assistant that learns player preferences to provide personalized gameplay guidance, rules explanations, and narrative suggestions during tabletop and digital game sessions. The system maintains a persistent player profile that accumulates knowledge about individual preferences, skill levels, and play patterns over time, adapting responses based on historical context rather than starting fresh each session.
Bandai Namco filed 1 patent for an AI system that creates diverse NPC behaviors by blending multiple players' operational habits rather than mimicking a single player. The approach uses weighted combinations of different players' behavioral data to exponentially increase NPC diversity beyond the number of available samples, making computer-controlled characters feel more unpredictable.
AT&T received 1 patent for an AI-powered content filtering system that dynamically personalizes gaming and media experiences by detecting and adapting mature content in real-time. Rather than blocking access entirely, the system intelligently modifies the content itself, adjusting visuals, audio, and gameplay mechanics based on individual user profiles and dynamically detected content severity levels.
Triumph Labs filed 1 patent for an AI-powered matchmaking system that dynamically adjusts skill-based matching algorithms during low player availability periods. The system uses predictive algorithms to preemptively modify matching criteria before players experience long wait times, balancing competitive fairness with player retention through real-time player liquidity analysis rather than fixed thresholds.
All data sourced from USPTO patent filings. Google Patents may take several weeks to index recent publications. If a link is unavailable, search for the patent number at USPTO Patent Public Search.