Cool AI Patents of the Month: Spotting Chatbots and Stopping Bullies

Welcome to the very first installment of Cool AI Patents of the Month. Each month, we’ll look at some of the more eye-catching and creative ways Artificial Intelligence (AI) is showing up in innovations that themselves wind up represented in patents or patent applications. The goal isn’t to get overly technical, but to highlight technology that’s genuinely cool and thought-provoking — the kind of inventions that make you say, “Wow, that’s clever.”

In this inaugural installment, we focus on how AI is advancing in numerous ways to address real world problems. The first patent we review is one from SAP that tackles the problem of knowing whether you’re chatting with a real person or an AI chatbot. And the second, a patent application, was filed by Tyco Fire & Security and uses AI to spot bullying as it happens.

Patent No. 12,438,836 – SAP: Human or Machine?

We’ve all been there — you’re typing away in a chat window and suddenly wonder, am I talking to a real person or is this an AI model pretending to be one?

SAP’s U.S. Patent No. 12,438,836 answers that question with an AI system of its own. The invention uses a cluster of machine learning models that evaluate a chat participant’s behavior and responses to determine whether they’re human. And while some might not find it necessary to know whether a chatbot is an AI pretending to be human, there is no doubt that technology that can decipher between an AI pretending to be human and a real human could at some point be quite impressive, and important.

In this case, the invention uses multiple models, which each look for different clues. One compares response time to message length, reasoning that long, instant responses often come from bots. Another tests knowledge and context, throwing in a few topical questions to see if the other side knows what’s going on in the real world. A third model asks the same question multiple times — a human might get annoyed, while an AI model will usually respond politely each time. Finally, the system checks how the other party reacts to being told they’re wrong. People tend to double down; chatbots tend to apologize.

Put it all together, and the system generates a score showing how likely it is you’re talking to a human. It’s smart, slightly ironic, and something definitely worth trying once it is deployed.

U.S. Patent Publication No. 2025/0316087 – Tyco Fire & Security: AI with a Heart

Next, we have a U.S. patent publication (not yet issued) from Tyco Fire & Security, and it’s one of those ideas that really shows AI’s potential for good. Their technology uses AI to detect bullying in real time — a tool that could make schools safer for students.

Here’s how it works: cameras capture live video streams, which are fed into an edge device running a 3D-enhanced neural network. The system analyzes body language, gestures, and interactions to spot bullying as it unfolds, then notifies school administrators so they can step in right away. And it is easy to envision this technology being deployed more widely as it matures so law enforcement could identify all kinds of potentially aggressive and violent interactions that are brewing before something actually happens.

What’s particularly clever is how the model scales traditional 2D networks into 3D form, which keeps computational costs low while maintaining accuracy — a big deal for devices running in real time.

Patents with Purpose

From detecting digital imposters to protecting kids in the schoolyard, these patents show just how wide the AI frontier has become. Each reflects innovation with both sophisticated engineering and a sense of purpose — and that’s exactly what makes them cool AI patents.