Redefining ‘Recording’ Isn’t an Option

With the rise of AI hardware, tech companies are redefining what it means to record. Traditionally, if a device had a microphone or camera, it was either actively recording or it wasn’t—clear-cut, right? Microphones pick up sound, cameras capture images, and that data is stored on a device, sometimes uploaded to the cloud for processing, ideally with strong encryption. This data represents a moment in time, making it a recording.

But that might be changing.

Recently, it was reported that a major tech company is developing a smart home camera that won’t save video recordings. Instead, it will use AI to analyze visual data and generate text snippets describing activities in your home. It’s said to include facial recognition capabilities, but without retaining any video footage, aiming to enhance privacy. This technology is reportedly similar to features expected in future wearable devices.

This concept echoes the Audio Intelligence features found in smartwatches, which have long included microphones primarily for voice commands and calls. Among the four Audio Intelligence features, two have sparked significant discussion: Live Rewind, which provides a transcript of the last 15 seconds, and Siri Recap, which summarizes daily conversations.

In contrast, smart glasses from another company can take photos and record video, storing that data on your phone. The rumored approach from the tech giant would mean that while cameras can see and microphones can hear, neither would actually save any data.

Does this still count as recording? While the company hasn’t commented on these rumors, its public stance on privacy suggests it doesn’t consider it recording. The company claims these features don’t retain audio, with raw data deleted after processing in a secure area of the device. Any cloud data is handled through private computing, inaccessible to the company or third parties.

This isn’t just a one-company issue. A senior director at another tech firm mentioned that they are also exploring this idea. What if smart glasses could process visual data without saving any footage? In this scenario, the camera would function merely as an image sensor.

By these definitions, something must be preserved and reviewable to be considered “recording.” But wouldn’t text transcripts from a smart home camera or smartwatch qualify as preserved data? Or are companies simply navigating legal definitions to make AI hardware more acceptable?

While it may seem trivial to debate semantics, there are significant concepts at play.

Regardless of personal views on privacy, it’s widely accepted that individuals have control over their own data. Before devices could “always listen,” they blended into public and private spaces more seamlessly. The latest gadgets, however, require users to be more vigilant about the privacy of others. Humans don’t exist in isolated bubbles; we interact in various environments—some private, others public, and many in between. The rules differ greatly from one setting to another, and even personal memories, while subjective, can’t be shared objectively like recordings can.

There’s also the challenge of signaling intent when there are no universally recognized standards. For instance, smart glasses from one company feature a white, pulsing light to indicate recording, while the common color for recording lights is red. This could confuse those unfamiliar with the technology. Additionally, these privacy indicators can be easily obscured, and not all smart devices are required to signal their recording capabilities in the same way.

The line between devices that can record and those that can’t is becoming increasingly blurred. If smart glasses start incorporating cameras that don’t save footage, distinguishing them from those that do will be challenging. If a device that lacks a microphone resembles one that records audio, how can the average user tell them apart?

Currently, the focus of the surveillance tech debate is primarily on hardware. The assumption is that removing cameras resolves the issue, but it doesn’t eliminate the potential for misuse. The discomfort often stems from the knowledge that a piece of data about you could exist on someone else’s device, with unknown future uses.

This tension surrounding AI hardware is significant. For AI to be effective and personal, it needs to understand you and your surroundings. It must “see” and “hear” your experiences, which implies a need to retain that information. Thus, in many ways, it must record aspects of your life. The motivations for tech companies to gather this data are clear: personal data translates to profit, and profit leads to influence. What remains uncertain is what society gains from this exchange—are we merely promised convenience?

Regardless, it’s evident that tech companies are committed to integrating AI into every device while also creating new forms. This trend is most visible in smartphones and wearables, but it will soon extend to smart homes, TVs, laptops, tablets, and headphones in ways consumers may not yet anticipate.

What we allow these AI devices to remember will reshape our culture. It’s crucial that we reach a societal agreement on what constitutes a recording and who, if anyone, has the authority to redefine that term.