Always on, always listening: New AI devices are pushing legal limits and could put users in jeopardy

A titanium Apple Watch Ultra 4 during a product unveiling event at Apple Park in Cupertino, California, US, on Wednesday, Sept. 9, 2026.

David Paul Morris | Bloomberg | Getty Images

Chris Thomas is all-in on AI gadgets.

Thomas, 43, wears a pair of Meta's camera-equipped glasses, a Fieldy note-taking pendant around his neck and a wristband from Amazon's Bee, all of which record his daily life. A Pittsburgh-based business owner and a livestreamer, Thomas relies on the devices to help him stay organized, pay bills on time and set reminders. 

"I'm forgetful AF," Thomas said in an interview. "My life is so crazy that not having to remember to remember is fantastic." 

Not everyone is on board with his lifestyle. Earlier this week, a bystander at a local bar noticed his Meta Glasses. 

"He was like, 'I don't feel comfortable,'" Thomas said. "I don't turn it off, I take it off completely because people don't really believe it's off." 

It's the type of interaction users are increasingly navigating as more tech companies introduce AI-powered, always-listening devices and features, which have been marketed as useful tools to transcribe meetings, keep track of conversations, set reminders and offer personalized recommendations. 

The latest example of these gadgets arrived Thursday with Amazon's Bee Frontier wearable and Bee Eclipse, a card-shaped device that attaches to a smartphone. Both devices are designed to record, transcribe and analyze a user's day.

Amazon's Bee Frontier device can be worn on a wrist or clipped on to clothing.

Amazon

Apple's latest Apple Watch Series 12, unveiled in September, will have Siri Recap, which "ambiently" takes notes throughout the day and generates summaries of conversations. Another feature called Live Rewind can recap 15 seconds of a conversation via a text transcription.  

Meta's popular smart glasses can take photos, capture video and record voice notes or reminders. Startups including Plaud and Pocket also developed card-sized note-taking devices, while Friend offers an AI necklace. 

The devices have fueled concerns that they could erode privacy, by enabling covert recording without obtaining bystanders' consent and collecting sensitive data that's potentially accessible to others. 

Amazon, Google and other makers of voice-activated smart speakers faced similar privacy concerns when their devices first hit the market, in some cases leading to lawsuits and regulatory fines. 

But this new generation of AI devices is also testing the limits of privacy law.

Unlike earlier smart speakers that were activated via a so-called "wake word" and sometimes stored recordings or sent audio to the cloud, some of these products can continuously listen, interpret speech, and extract its meaning without retaining the underlying audio. This capability has raised a largely untested question about whether that processing itself can amount to an unlawful interception.

"With new technology like this, my statement is always to embrace with caution," Karen Panetta, dean of graduate education at Tufts' School of Engineering and a fellow at the Institute of Electrical and Electronics Engineers, said. 

She added, "If I'm giving permission for myself to be recorded, or for everything I look at to be recorded, that doesn't mean everybody else around you has given that consent."

Apple and Amazon appear to have sidestepped some of the privacy concerns that dogged earlier voice assistants. 

Apple says the feature does not store audio, send recordings to its servers, or involve human review. Amazon has said it doesn't store audio from Bee devices, and conversations are processed in a secure environment "where no one, including Amazon, can access your data." 

But that may not resolve the legal question.

Many state wiretap and eavesdropping laws are written around whether a private communication was intercepted or its contents were acquired — not simply whether a conventional audio recording was created.

Colin Zick, a privacy and data security attorney at Foley Hoag, said Apple appears to have learned from the first wave of litigation over voice assistants. But the new technology raises questions those earlier cases never had to answer: Can a system that is continuously listening, interpreting speech and turning it into a summary still amount to an interception if it never saves the underlying audio?

Meta CEO Mark Zuckerberg wears the Meta Ray-Ban Display glasses, as he delivers a speech presenting the new line of smart glasses, during the Meta Connect event at the company's headquarters in Menlo Park, California, U.S., Sept. 17, 2025.

Carlos Barria | Reuters

Some state laws define interception broadly. Massachusetts, for example, defines an intercepting device to include equipment capable of "transmitting, receiving, amplifying, or recording" a communication, while the "contents" of a communication can include its substance and meaning. 

In Zick's view, a device that continuously receives speech and produces a written summary could potentially fall within that language even without retaining an audio file.

And in some states, the bigger legal risk may fall not on Apple, Amazon or other AI gadget makers but on the person wearing the device.

"A non-wearer did not buy the watch, agree to Apple's terms, or choose to have their conversation summarized in the app," Zick said.

Read more CNBC tech news

Daniel Pi, an associate professor at the University of New Hampshire Franklin Pierce School of Law, said New Hampshire's statute is broad enough that he believes the Watch's new audio-intelligence features would qualify as an interception. The law covers not only recordings, but the "aural or other acquisition" of the contents of a communication.

That could create an unusual situation in which an ordinary consumer turns on a feature built into a mainstream device and, at least technically, runs afoul of state law by using it during a private conversation without everyone's consent.

Pi said a user in New Hampshire could potentially face civil or criminal liability for doing so, though he stressed that the practical risk of prosecution is far lower. It could be difficult to know whether someone had the feature enabled in the first place, and he said he would be surprised to see prosecutors pursue cases involving a bare statutory violation where no one was harmed.

The legal questions are also different from those raised by earlier smart speakers and voice assistants. Those cases often centered on devices mistakenly activating or capturing speech that was not intended as a command. Here, the feature is designed to acquire and interpret the substance of a conversation — and the user has deliberately opted in.

That changes the analysis, Pi said, and could also shift responsibility toward the wearer rather than the technology company itself.

The broader privacy concern extends beyond who is legally responsible.

Michael Stokes, a former CIA officer and senior vice president of emerging technologies at Veilant, said the bigger shift is that people can be swept into AI-powered surveillance without owning or knowingly interacting with the device doing the collecting.

"I think the bystander problem is really the concern I have — that you can be surveilled without owning the technology," Stokes said.

He pointed to the growing number of cameras, microphones, and sensors embedded in consumer devices, from smart glasses to watches and AI recorders. The difference now, he said, is that AI can rapidly interpret what those devices collect and draw meaning from far larger volumes of information than earlier generations of consumer technology.

For Stokes, that makes the Apple Watch part of a much larger change in the privacy landscape. During his time at the CIA, he helped develop the concept of ubiquitous technical surveillance, describing a world in which everyday devices continuously generate information that can reveal identities, locations, and associations. 

The concern is not simply that more data can be collected, he said, but that the cost of making sense of it has fallen dramatically.

In the past, Stokes said, intelligence collection could require thousands of hours of work to gather, sort and analyze information about a person. AI increasingly automates that process, making it cheaper and easier to extract useful information from data that may have been gathered incidentally.

That creates a new kind of exposure for people who never opted into the technology in the first place — whether it is a private conversation picked up by a nearby wearable or information collected by a device they do not own.

"We no longer have these pockets of small privacy that we thought we did in the past," Stokes said.

Aida Baradari, founder of the startup Deveillance, is trying to give some control back to people worried about being recorded without their consent.

The company created a hockey puck-shaped gadget, called the Spectre I, that uses AI to scan for nearby devices and sends out "adversarial signals" that scramble a person's speech before microphones can process it.

Baradari said she watched as more AI wearables hit the market and felt users should have a greater say in how they interact with them.

"We need to be able to use technology while not giving up our core values or rights," she said.