The danger from 'always listening' AI devices
Chris Thomas, 43 years old in Pittsburgh, Pennsylvania (USA), completely trusts devices that integrate artificial intelligence. He always wears glasses with a camera, a pendant with recording and note-making features, and a smart bracelet. He relies on them to organize his work, pay bills on time, set daily reminders, and monitor his health.
"I'm extremely forgetful," he explained to CNBC. "Life is so busy that I can't remember everything."
Thomas, who runs a business and is a content creator, said not everyone likes his lifestyle. "A passerby looked at the Meta glasses I was wearing and said: 'I don't feel comfortable. I immediately took them off instead of just turning off the recording feature, because maybe people didn't really believe it was turned off,'" he said.
People like Thomas are increasingly easy to meet in real life. In the context of technology companies continuously launching devices advertised with AI integration, which can automatically record meetings, monitor conversations, set reminders and make personalized recommendations... at accessible prices, the number of users is also increasing.
Citing data from research firm Omdia, Reuters said the number of AI-integrated eyeglasses shipped globally has skyrocketed to 8.4 million units in 2025, compared to 2.1 million in 2024.
In addition, the Apple Watch Series 12 launched in September has the Siri Recap function to help take notes and summarize conversations, along with Live Rewind summarizing 15 seconds of conversations via text. Amazon mid-last week also introduced Bee Frontier and Bee Eclipse, card-like devices that can record, transcribe and analyze users' daily activities.
Many smaller companies like Plaud and Pocket develop card-sized note-taking devices, while Friend offers necklaces with built-in artificial intelligence. The explosion of AI helps personalized products for users become more and more popular.
However, these devices also raise concerns about the risk of privacy invasion. Unlike previous products, which needed to be activated with a "keyword" or sometimes store and send audio to the cloud, the new device can continuously listen, interpret voices and extract meaning using artificial intelligence.
NYPost noted that on social networks, many people expressed discomfort when smart devices record data themselves. Three Reddit threads about the Meta Rayban glasses currently have more than a thousand comments, reflecting the recent software update that allows the device to start recording without saying "Hey Meta" - the keyword used to wake up from sleep mode. Some even said they were planning to file a class action lawsuit against Mark Zuckerberg's company. Meta later confirmed "the issue has been resolved", but did not mention details.
"With new technologies like this, my view is always that they need to be approached with caution," said Karen Panetta, dean of graduate education at Tufts School of Engineering and a member of the Institute of Electrical and Electronics Engineers (IEEE). "If it's okay to record myself or everything I look at, that doesn't mean everyone around me has agreed."
Some other experts are more concerned, arguing that privacy rights are not limited to who is legally responsible. "I think the real concern lies with the people around them: they can be monitored without owning any technological devices," said Michael Stokes, a former CIA officer who is now senior vice president of emerging technology at Veilant.
According to Stokes, the increasing number of cameras, microphones and sensors integrated into consumer electronics devices makes things difficult to control. The difference compared to before is also the presence of AI - technology that can quickly interpret what is collected and draw meaning from a "much larger" volume of information.
Some technology companies are trying to avoid the privacy concerns that existed with previous devices and voice assistants. For example, Apple confirms that the notes feature on Apple Watch does not store audio or send recordings to the server. Amazon confirmed the same thing on the newly launched Bee device duo.
Colin Zick, a privacy and data security lawyer with the law firm Foley Hoag, assessed that Apple seems to have learned from the wave of lawsuits related to voice assistants. However, new technology is raising a series of questions that previous lawsuits have not had to answer: Can a system that continuously listens, interprets voices and summarizes them be considered eavesdropping, when it has deleted itself and not saved the original recording?
Besides, one of the difficult problems to solve with continuous listening technology is the privacy of people around. When a person wears an AI bracelet or smart glasses, the people participating in the conversation may not know the device is active if they are not paying attention, making it difficult to know whether the audio is processed locally or sent to the server. They also do not have access to settings to check the data that has been collected or request deletion of information.
On October 13, the European Data Protection Board (EDPB) warned that devices that look like regular glasses but have cameras, microphones and Internet connections can cause people to have information collected without their knowledge. Norway is one of the governments with a tough stance on AI glasses, announcing in early October that it intended to ban the use of the device in some public places.
"There is no single off button that is enough to control," Dr. Roman Yampolskiy, a computer science instructor at the University of Louisville, told NYPost, adding that production companies need to install a recording warning light or sound "loud enough to be seen" on the device, add automatic facial blurring in the recording software, and if users commit abuse, they will receive "adequate punishment."