Tech firms are quietly redefining what counts as 'recording' — and we need to push back

Traditionally, it has been simple enough to know when a device is recording: cameras capture images, microphones capture sound, and that data gets stored somewhere. But Apple, Google, and Meta are introducing a twist with their latest AI-powered hardware, proposing that devices can "see" and "hear" without technically recording in the old sense.

The new argument#

Last week, Bloomberg reported that Apple is developing a smart home camera that won't store video at all. Instead, AI will process the visual feed and output only text summaries—"it's 3 p.m. and your cat is napping on the couch"—while discarding the original footage. The same principle applies to rumored cameras in upcoming AirPods and to Apple's Audio Intelligence features on the latest Apple Watch models, which transcribe and summarize audio but claim to delete the raw sound after processing.

Google's wearables chief recently sketched a similar vision to The Verge: imagine smart glasses with cameras that don't record or store any video footage at all, just perform real-time AI analysis on what they see. By these definitions, nothing is preserved long-term, so technically nothing was "recorded."

Why this distinction matters—and fails#

The logic is appealing on the surface. If no footage sticks around, how can it be misused? But the argument collapses under scrutiny. Text transcripts generated by these systems—summaries of conversations, descriptions of your home, or AI-derived insights about your daily routine—are themselves preserved data. They're reviewable. They can be subpoenaed. In what meaningful way are they different from a video file?

What's really happening is a sleight of hand: tech companies are trying to reshape our understanding of "recording" to exclude the raw data while ignoring that the processed output is still a preserved artifact.

The privacy gaps that remain#

Beyond semantics, there's a deeper problem. When you wear an always-listening camera or microphone, you're no longer the sole custodian of your privacy—you're forcing it onto everyone around you. Your partner having a private conversation in your home, your coworker's face in a coffee shop, a stranger's voice in a public restroom: all of this enters these devices and gets analyzed by AI. You don't consent for them; they have no choice.

Physical signaling is already failing. Meta's smart glasses use a white pulsing light to indicate recording, but most people expect a red light and won't notice the signal anyway—especially if it's obscured by hair or sunlight. As more companies add always-on cameras to glasses, rings, and pendants, the visual cues will become even more inconsistent or disappear entirely. How will anyone know what's watching them?