Microsoft

The 3 Platforms patents each tackle a distinct friction point in multiplayer and assisted gaming. One describes a proximity-based login system where a player's smartphone continuously verifies their physical location near a console using WiFi signatures, IoT scans, and GPS data, automatically authenticating their profile when they arrive at a friend's home and terminating the session the moment they leave. Another addresses the social dynamics of in-game coaching by monitoring interactions between helpers and younger players in real time, restricting how helpers can communicate or engage based on the player's age rather than relying on static content filters. The third rounds out the set by tracking whether a player completed a game achievement on their own or with assistance, attaching a formal label to each accomplishment so the record reflects how it was actually earned.

Across 7 patents, the AI & Machine Learning category covers a wide range of applications, though several cluster around the problem of in-game assistance. One filing describes a system that analyzes aggregate gameplay data to detect where players are collectively struggling and then proactively delivers contextual help before players even ask for it. A related patent takes this further by training machine learning models on expert gameplay footage, producing AI assistants capable of actively playing through difficult sections in real time rather than pointing players toward static guides. A third filing in this space lets helpers take control during coaching sessions but filters their inputs based on current game state, blocking actions that would spoil story moments or bypass intended challenges. Beyond assistance, Microsoft also filed a patent for a neural network that generates physically plausible hand poses for VR, AR, and robotics training data by learning the biomechanical constraints of the human hand rather than relying on manually engineered rules. 2 patents address voice transformation: one analyzes a player's appearance and their character's physical attributes to automatically blend appropriate voice characteristics in a role-playing context, while a second builds on that concept by updating the voice output every frame based on game state, so a character sounds winded while sprinting or muffled while submerged. The final AI & Machine Learning filing covers photorealistic avatar generation using Gaussian splatting constrained by a statistical prior model trained on human appearance data, allowing high-fidelity digital twins to be produced even from sparse or low-quality input.

The single Game Engines patent describes a testing Platforms that uses intelligent agents to interact with games the same way a human player would, sending inputs through external devices rather than embedding any code inside the game itself. Because the agents operate from the outside in, the same testing solution works across any game engine or Platforms without requiring custom integration work for each new title.

Cloud gaming contributes 1 patent, centered on a remote assistance system that allows an expert player to step into a struggling player's session without traditional co-op infrastructure. The system extracts the player's saved game state, loads it into a separate cloud session where a helper can take over and complete the difficult section, and then writes the updated state back into the original player's game when the helper is done.

The 1 Audio patent extends precomputed wave-based acoustic modeling into environments that change while a game is running, specifically targeting doors and other openings whose states shift dynamically at runtime. Traditional precomputed approaches break down when scenes are not fully static, so this system stores portal path data separately from ambient acoustic parameters, allowing the Audio model to update cheaply whenever an opening changes state without recalculating the full scene.