PULSE
AI 认知观点 VIEWPOINT重要度 62

AI硬件正在重新定义“录音录像”

The Verge AI · 2026-10-06 · 原题《We can’t just change the definition of ‘recording’》 · 原文约 6 分钟 · 阅读原文 ↗
AI硬件重新定义“录音录像” 科技公司:不保存原始音视频 ≠ 记录 传统定义 · 边界清晰 设备要么在录,要么没录 数据被保存 → 就是记录 可回看、可传唤 规则统一,公众可理解 AI介入 新定义 · 模糊地带 不保存原始音视频 AI生成文字摘要/转录 文字同样可保存、可回看 利用法律“可传唤”漏洞? 核心质疑:什么才算“记录”? 苹果 摄像头只生成文字 Watch转录摘要 谷歌 眼镜只做传感器 AI处理不存画面 Meta 白色脉冲灯提示 易被遮挡/误解
科技公司正把“不保存原始音视频”包装成不算记录,但生成的文字摘要同样是可保存、可调取的记录,社会需要就“什么算记录”达成共识。

论证要点

  1. 传统定义:带麦克风或摄像头的设备要么在录要么没录,数据被保存下来就是记录,边界清晰。
  2. 苹果传闻中的智能家居摄像头不保存视频,只用AI生成描述性文字并支持人脸识别;Apple Watch的Live Rewind和Siri Recap也生成转录和摘要,苹果称这些功能不保存音频。
  3. 谷歌可穿戴设备负责人也提出类似设想:智能眼镜的摄像头只做图像传感器,用AI处理视觉数据但不保存画面。
  4. 按这些定义,只有被保存且可回看才算“记录”,但AI生成的文字转录和摘要同样可保存、可回看,作者质疑这只是在利用法律上什么可被传唤的漏洞。
  5. 新设备要求用户主动管理他人隐私,而人类生活在公私模糊的空间中,餐厅、剧院、办公室、公共厕所的规则各不相同,缺乏统一规范。
  6. Meta智能眼镜用白色脉冲灯提示录像,但通用录像灯是红色,普通人可能不懂;灯还容易被刘海或阳光遮挡,其他AI硬件甚至可能完全没有提示灯。

对实践的启示

做AI硬件和应用的人需要意识到,“不保存原始数据”不等于没有记录,生成的文字摘要同样敏感且可被调取。产品设计应明确提示记录状态、统一信号规范,并提前考虑他人隐私和合规风险。

作者承认的反方与局限

作者承认,去掉摄像头或麦克风并不能解决监控问题,但会让技术更难被用于恶意目的;同时,苹果等公司声称不保存原始数据、通过安全隔离和私有云计算保护隐私,这些说法本身也是一种立场。

原文摘引

By these definitions, something has to be preserved and reviewable to be “recording” you. But wouldn’t any text transcripts generated by a smart home camera or smartwatch count as preserved and reviewable?
The queasy feeling comes from knowing a preserved snippet of you can exist in someone else’s device with no knowledge of how it’ll be used.
What we allow these AI gadgets to remember will rewrite culture as we know it.
▸ 展开原文全文(19 段,抓取自 The Verge AI)

With AI hardware, tech companies are pushing the definition of what does and doesn’t constitute a recording. For most of gadget history, it’d be reasonable to assume that a device with a microphone or camera is either recording you or it isn’t; it’s either on or off, without much gray area. Microphones capture sound. Cameras capture images. That data is saved onto a device. Sometimes, it’s also sent to the cloud — hopefully with end-to-end encryption — for processing. That data is a preserved instant in time and is therefore a recording.

Perhaps not anymore.

Last week, Bloomberg’s Mark Gurman reported that Apple is working on a smart home camera that doesn’t save video recordings. Instead, AI uses the visual data to create text snippets that describe what’s happening around your home. Supposedly, it’ll support facial recognition, but without video footage. The idea is that it still protects some privacy. This tech, Gurman says, is similar to how the cameras in rumored future AirPods will also work.

It’s similar to Apple’s Audio Intelligence on the Apple Watch Series 12 and Apple Watch Ultra 4. The Apple Watch has had a microphone for ages, mostly for Siri and taking phone calls from the wrist. Of the four Audio Intelligence features, two have caused the most debate online: Live Rewind and Siri Recap. The former produces a transcript of the past 15 seconds, whereas the latter can generate high-level summaries of the conversations you have throughout the day.

[Image: All Audio Intelligence features are opt-in. As you can see in this photo, Apple states in its Music Detection language that it “does not record audio.” But once a song is detected, is that not a record that you heard that song? https://platform.theverge.com/wp-content/uploads/sites/2/2026/09/268739_Apple_Watch_Ultra_4_AKrales_0212.jpg?quality=90&strip=all]

In other words, Ray-Ban Meta glasses take photos and record video, which is stored on your phone. This rumored Apple approach would mean cameras can see you, microphones can hear you — but neither is “saving” anything.

Does this count as recording? Apple doesn’t comment on rumors, but based on how it’s publicly communicated its Audio Intelligence privacy philosophy, I’d argue the company doesn’t think so. Apple says these features don’t save any audio. The raw data is deleted once it’s finished processing in a separate, secure exclave on the S11 processor. Any information sent to the cloud is through private cloud compute and is therefore inaccessible to Apple or third parties.

It’s not just Apple, either. Sandeep Waraich, Google’s senior director of wearables, told The Verge in a conversation that Google is also mulling this concept. What if, he posited, cameras on smart glasses didn’t record data? For example, if the glasses could process visual data in your environment using AI, but didn’t save any footage for posterity. In that way, Waraich argued, the camera could “just” be an image sensor.

By these definitions, something has to be preserved and reviewable to be “recording” you. But wouldn’t any text transcripts generated by a smart home camera or smartwatch count as preserved and reviewable? Or, in trying to make AI hardware more palatable, are tech companies relying on what can be subpoenaed in court?

It might seem silly to argue semantics, but multiple concepts are being tested here.

Whatever your personal opinions about privacy, it’s generally agreed that you can do whatever you want with your own data. Before gadgets could “always listen,” they slotted into public and private spaces more seamlessly. This new crop of gadgets, however, asks you to be a more active steward of other people’s privacy. After all, humans don’t live in hermetically sealed bubbles. We live, work, and hang out in a multitude of environments — some private, others public, and plenty in a gray area. The rules for a restaurant differ from those of a theater, office, or sporting venue. They’re vastly different from a public restroom, even though these are all at least semi-public spaces. Even if you “record” everything you see in your mind, human memories can’t be shared objectively. They’re colored by biases and warped by perspective. Recordings are less subjective and arguably factual.

There’s also the question of signaling intent in the absence of universally understood norms. Take Meta’s smart glasses. They use a white, pulsing “privacy light” to indicate when someone’s recording video. Conversely, the universal color for recording lights is red, so a less tech-savvy person may not even know what a white light signifies. (These are, after all, still niche products.) Even without modification, white privacy lights are easily obscured by bangs and a bright sunny day. This is also a new norm; other smart glasses or AI hardware companies aren’t obligated to signal this intent in the same way as Meta. An AI ring or pendant that records audio or video may have no light at all.

[Image: Would you notice this privacy light in person? https://platform.theverge.com/wp-content/uploads/sites/2/2025/10/257980_Meta_Ray-Ban_Display_AKrales_0356.jpg?quality=90&strip=all]

Even more confusing, products that can record increasingly look like products that can’t. That could be exacerbated even further if tech companies begin adding cameras that don’t record visual footage. If we start adding cameras to smart glasses that don’t record video footage, how will you tell them apart from pairs that do? If an Oura Ring — which has no microphone — looks like a Vocci Ring — which does — and both look like regular rings, what is the average person supposed to do?

Right now, hardware seems to be the main focus of the surveillance tech debate. There’s a sense that getting rid of cameras solves the problem. It doesn’t, though it makes it harder to use this tech in nefarious ways. But arguably, it’s not the camera or microphone that gives people the ick. The queasy feeling comes from knowing a preserved snippet of you can exist in someone else’s device with no knowledge of how it’ll be used.

This is where all the tension surrounding AI hardware converges. For AI to be “good” and truly personal in the way Big Tech imagines, it has to know a lot about you — and the people in your life. It has to “see” what you see, hear what you hear, and learn from your day-to-day actions. To do that, it has to “remember” these things. So in some ways, it must record your life and everything that happens in it. The incentives for Big Tech to collect this data are obvious: Personal data is money, money is power, and power is influence. What’s less clear is what society gets from this tradeoff. Vague promises of helpfulness and convenience?

[Image: This Vocci Ring can record conversations and its privacy light is extremely hard to notice. https://platform.theverge.com/wp-content/uploads/sites/2/2026/05/268537_Vocci_smart_ring_AKrales_0123.jpg?quality=90&strip=all]

Regardless, it’s clear tech companies have decided to barrel forward with stuffing AI into every existing gadget while inventing new form factors. Right now, it’s most clear in phones and wearables, but these considerations are soon headed to the smart home, TVs, laptops, tablets, and headphones. Likely in ways consumers have yet to consider.

What we allow these AI gadgets to remember will rewrite culture as we know it. This is why we must reach a societal consensus on what constitutes a recording — and who, if anyone, gets to change that definition.