A new category of consumer hardware is emerging, and its defining feature is not a screen, a camera, or a speaker. It is a microphone that never stops listening. Amazon’s Bee subsidiary recently introduced a wristband called the Bee Frontier, priced at $75, designed to record conversations and generate transcripts and summaries accessible on a phone or computer. The device looks like a screenless fitness tracker. Its purpose is to solve forgetfulness. That framing, modest and practical as it sounds, obscures a much larger shift in how public and private life may work going forward.
A Market Built Around Human Memory’s Limits
Amazon is not alone in this space. Apple introduced a feature called Siri Recap for recent Apple Watch models, capable of summarizing conversations the watch listens to. A brand called Vonder is building smart glasses that track the wearer’s side of every conversation. Workplace-focused recording devices are already being marketed to offices. The pitch across all these products is consistent: humans forget things, and AI can remember them instead.
This is a genuine problem being addressed. Details from meetings slip away, phone calls blur together, and the cognitive load of modern work is real. The question is not whether the problem exists. The question is what gets traded away in solving it.
Amazon has made visible efforts to address the most obvious concern. Audio captured by Bee devices is processed in real time and then deleted. Transcripts and summaries are stored in an encrypted environment that Amazon’s own employees reportedly cannot access. An LED light activates whenever the device is recording. These are meaningful design choices, and they reflect awareness that the product operates in sensitive territory.
The LED Light Problem, and What Lies Beyond It
Here is what most coverage of these devices misses: the LED indicator, the most common privacy signal across AI wearables, is a weak solution to a structural problem.
Sara Geoghegan, senior counsel at the Electronic Privacy Information Center, has pointed out that an LED light may be visible to some bystanders but not recognizable to all of them as a signal that recording is active. It is small. People may not notice it. In states like California and Florida, where two-party consent laws require all parties in a private conversation to agree to being recorded, a blinking light may not constitute legal consent.
The problem extends further. Some users of Meta’s Ray-Ban AI glasses have reportedly tampered with their LEDs to record others without their knowledge, which led Meta to remotely disable thousands of pairs of glasses. This is not a hypothetical risk. It has already happened.
Some manufacturers have tried to go further than LEDs. HTC’s Vive Eagle smart glasses stop recording when a bystander says “stop recording.” Limitless, an AI pendant maker, required audible consent from bystanders before capturing their voices when it launched its device in 2024. Limitless has since been acquired by Meta, which has not carried that consent feature into its own products.
Meta’s own camera-free AI glasses present a different kind of challenge. They have no LED at all, making them visually indistinguishable from ordinary glasses. The translation feature on Meta’s Ray-Bans keeps the LED off while active, and Meta’s servers may retain transcripts of translated conversations for up to a year unless users manually delete them.
What Pervasive Recording Does to Public Space
The deeper concern, raised by Oxford University professor Viktor Mayer-Schönberger, is not about any single conversation being captured. It is about what disappears when every conversation could be captured.
Public spaces have historically offered a kind of built-in forgetting. A bad joke, a mispronounced word, an offhand remark: these things faded. The people who heard them moved on. That ephemerality was not a flaw in public life. It was a feature. It allowed people to be imperfect, spontaneous, and unguarded in ways that are fundamental to how humans actually interact.
Mayer-Schönberger argues that this ephemerality has already begun to evaporate. If every word spoken in a bar, a museum, or a workplace might be preserved and analyzed indefinitely, the rational response is to speak less freely. He draws a direct parallel to behavior in authoritarian societies, where people self-censor in public out of awareness that they are being observed. In liberal democracies, he argues, participation in public life is foundational to how society functions. A retreat from that participation, driven not by government surveillance but by consumer hardware, would be a serious loss.
The path forward, in his view, is not a technical fix but a social negotiation. Societies have renegotiated public behavior before: audiences once shouted at actors during theatrical performances, a practice now considered unthinkable. Norms around phone use in social settings shifted within a generation. The norms around AI recording are not yet settled, and that is precisely the moment to shape them deliberately. Without that process, Mayer-Schönberger warns, regulation will arrive only after something goes badly wrong, and it will likely be blunt and disproportionate.
In Short
AI wearables designed to record and summarize conversations are already on the market, priced accessibly, and backed by major technology companies. The privacy mechanisms built into these devices, primarily LED indicators and encryption, address some concerns but leave others unresolved, particularly around consent in crowded or public settings. The broader issue is not surveillance in the traditional sense. It is the gradual erosion of the informal forgetting that public life has always depended on. How societies choose to negotiate the boundaries of AI-powered memory will matter far more than any single product launch.
Based on reporting from Fast Company.