IBM Patents Software That Uses Cooperating AI Models to Spot Sensitive Data
Instead of trusting a single AI model to spot sensitive information, IBM's new patent describes a panel of AI models that check…
43 filings · category
AI safety patents
Watermarks, guardrails, moderation, and model protection: the filings where Big Tech works out how to keep AI honest, and how to prove what's fake.
Recent filings
43 total
Google's New Patent Makes Private AI Training Far Less Expensive to Run
Training an AI model while keeping individual people's data genuinely private is expensive, and Google thinks it's found a way to make…
Google's New Patent Stops Its Smart Speakers From Talking to Themselves
If you've ever watched a Google Nest speaker respond to something it just said out loud, you've found the bug this patent…
Adobe Patents a System That Invents Fake Users to Train Its AI Assistants
Training an AI assistant requires thousands of example conversations, but real user data is hard to collect and raises privacy concerns. Adobe's…
Google Patents an AI Training Method That Protects User Data Without Sacrificing Accuracy
Training a recommendation AI on real user data is a privacy minefield. Google's new patent describes a way to do it with…
Amazon Patents a Targeted System for Keeping AI Image Generators Clean
Most AI image safety filters work like a sledgehammer — block enough and you avoid bad outputs, but you also wreck perfectly…
Qualcomm Patents Technology That Catches Its Own AI Errors to Speed Up Responses
Qualcomm has filed a patent for a technique that lets a single AI model draft its own answer in advance, check how…
Salesforce Patents a Way to Train AI Assistants to Refuse Unanswerable Questions
Most AI assistants will confidently make up an answer rather than admit they don't know — Salesforce is filing a patent for…
Google's New Patent Trains AI on Your Data Without Ever Seeing the Full Picture
Google is patenting a way to train AI models on your personal data without any single computer — not even Google's own…
OpenAI Patents a Way to Make Its AI Actually Think Before Refusing You
Most AI guardrails are more like a lookup table than a conscience — the model pattern-matches your request against a list of…
Google Patents a Way to Run Safety Drills on Its Own AI — No Humans Required
Google is patenting a pipeline that automatically fires thousands of test prompts at a generative AI model — text, images, audio, video…
IBM Patents a Risk-Scoring Guardrail That Stops AI Agents Before They Do Something Costly
AI agents that autonomously run business tasks are only as safe as the guardrails around them. IBM's new patent describes a risk-scoring…
Get the weekly Patentlyze email
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.