AI devices challenge what companies can claim counts as recording
Tech companies including Apple and Google are redefining what constitutes a "recording" as they develop new AI hardware. Rather than saving raw video or audio data, these devices would process information using AI to generate text descriptions, then delete the original data, raising fundamental questions about what companies can legitimately claim they don't "record" even when devices capture personal information.
Apple is reportedly developing a smart home camera that doesn't save video but creates AI-generated text descriptions and supports facial recognition, mirroring Audio Intelligence features on the Apple Watch Series 12, such as Live Rewind and Siri Recap. Google's wearables division is exploring similar concepts where cameras could process visual data without retaining footage. The debate centres on whether preserved text transcripts constitute a "recording," and raises privacy concerns in shared spaces where multiple people may be captured by these devices.
- Tech companies argue devices that delete raw data aren't "recording" anything.
- Smart cameras could transcribe conversations without saving audio files.
- Privacy concerns loom when devices capture people in shared spaces unknowingly.
New here? Start with this
Tech companies are developing new artificial intelligence devices that work differently from traditional cameras or microphones. Rather than saving video or audio footage, these devices analyse what they capture using AI to create text descriptions – such as identifying people or objects – then delete the original data. This raises a seemingly simple but consequential question: if a device captures information but doesn't keep the recording, has it actually "recorded" anything?
Apple is reportedly building a smart home camera that would work this way, creating descriptions of what it sees without storing the video. Google's wearables division is exploring similar technology for smartwatches and other devices. The approach mirrors features already available on some Apple products, such as the Watch Series 12, which can summarise conversations or sounds without saving the raw audio.
The distinction matters because people's understanding of "recording" influences how they think about privacy. If companies can argue they don't record anything – only store AI-generated summaries – they might claim different privacy protections than devices that keep raw footage. Yet these devices still capture sensitive information about people in shared spaces, including faces and conversations, which raises questions about whether the difference between capturing data and recording it is meaningful for privacy protection.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Tech companies argue that processing captured information through AI and deleting the raw data differs meaningfully from traditional recording. If video or audio is converted to text summaries and the original is discarded, they contend no persistent "recording" exists—only derived information. They suggest this approach actually enhances privacy by preventing long-term storage whilst retaining useful functionality, and that users selecting these devices understand what data persists.
The case against
Privacy advocates contend that capturing any personal information constitutes recording regardless of whether raw files are deleted afterwards. They argue the semantic distinction between raw data and processed summaries is meaningless to users whose private moments were captured without consent, particularly in shared spaces. Redefining "recording" based on technical processing methods, they warn, sets a dangerous precedent that could allow companies to evade privacy protections through procedural claims.
Read the full article at the source →
Originally published by The Verge as “We can’t just change the definition of ‘recording’”.