/Patent Tracker
GooglePatent applicationAI & softwareInterest 6/10

Google patents fast safety check for AI-generated text

Google describes a system that uses a smaller, faster AI model to check if the answer from a larger AI model is safe before showing it to a user. This safety filter is trained on examples of AI responses labeled as safe or unsafe to quickly block inappropriate content.

Why it mattersThis approach aims to reduce the delay users experience when AI systems filter their responses, making the interaction feel more immediate.

In the filing’s own words

Aspects of the disclosed technology include computer-implemented systems and methods for moderating generative content produced by machine-learned generative models. The system can obtain generative content produced by a machine-learned generative model in response to a user query and determine with a machine-learned safety model a safety classification for the user query. The machine-learned safety filter model includes a transformer and has a number of parameters that is less than a number of parameters of the machine-learned content classification model. The machine-learned safety filter model is trained on a set of input-output pairs with each input-output pair including an input comprising a generative model response and an output comprising a safety label.

What exactly they claim (claim 1)

1. A computer-implemented method, comprising: obtaining, by one or more processors, generative content produced at least in part by a machine-learned generative model in response to a user query; determining, by the one or more processors with a machine-learned safety filter model in response to the user query, a safety classification for the user query, the machine-learned safety filter model including a transformer and having a number of parameters that is less than a number of parameters of the machine-learned generative model, the machine-learned safety filter model having been trained on a set of input-output pairs, each input-output pair including an input comprising a generative model response and an output comprising a safety label; in response to the safety classification satisfying one or more criteria, providing, by the one or more processors, a response to the user query, wherein the response includes the generative content; and in response to the safety classification failing to satisfy the one or more criteria, filtering, by the one or more processors, the generative content from the response provided to the user query.

Facts

Company
Google (Google LLC)
Official title
Low-Latency Safety Filtering for Machine-Learned Models
Publication number
US20260289249A1
Application number
19084531
Filed
Mar 19, 2025
Published
Sep 24, 2026
Inventors
8
Claims
20

Classified under

G06N 3/0475 Physics & computingG06F 16/285 Physics & computingG06N 3/0455 Physics & computingG06N 3/09 Physics & computing
Read the full filing (official PDF) ↗ Open in the Patent Tracker

More from Google in AI & software

Summaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.