Will Siri Recap make the Apple Watch creepy? Privacy experts weigh in

I’ve always thought of the Apple Watch as one of the friendliest and most helpful Apple devices, sitting on your wrist with a colorful watch face, telling the time, nudging you to get off your tush, gently tapping you (or not) with text alerts, and even sounding the alarm if it detects anything amiss with your heartbeat.

But a new Siri AI feature, announced earlier this week during Apple’s fall hardware event, could land unsuspecting Apple Watch users who enable the option into the middle of a heated AI privacy debate, according to several experts I spoke with.

Welcome to another edition of Prompt Mode, your weekly AI newsletter.

I’m your host, Ben Patterson. Each week on Prompt Mode, I’ll be serving up analysis of the AI trends that matter to everyday users like you and me. Stay tuned for practical AI tips, hands-on experiences with the latest AI tools, and–you guessed it–prompts to help you get the most out of your AI assistants.

Thanks for reading, and if you like what you see, just sign up right here.

The feature, called Siri Recap, lets the new Apple Watch 12 and Apple Watch Ultra 4 keep an ear out for nearby conversations and deliver “high-level summaries” of what was said. Siri Recap will arrive later this year as a beta product, Apple says.

Alongside Siri Recap, the latest Apple Watches will offer Live Rewind, an on-demand feature that gives you a “text snippet” of the prior 15 seconds of a conversation. You activate Live Rewind by double-pressing the digital crown of your Apple Watch, an action that also triggers an audible chime as well as a telltale animation on the screen.

But unlike Live Rewind, Siri Recap can work passively in the background, intercepting nearby conversations whenever it’s on — which, potentially, could be all the time. Apple says Siri Recap users can choose “exactly where and when Siri Recap takes notes” via a schedule (such as “only at work” or “never at night”) and that the feature can be manually turned on or off from the pop-up Control Center.

Crucially, Apple says that Siri Recap doesn’t store recordings or verbatim transcripts of conversations, but instead “creates high-level Apple Intelligence-generated summaries” intended to “jog your memory” rather than capture everything that was said. In another key caveat, Siri Recap doesn’t track who said what, thus avoiding the use of voice fingerprinting technology.

Apple

Apple has taken other Siri Recap privacy measures, including insuring that financial data and “certain personal identifiers” are “omitted” from the summaries, as well as detailing a “Secure Enclave” in the Apple Watch’s new S11 chip that privately processes audio on the watch itself, while Privacy Cloud Compute handles cloud processing duties in an end-to-end encrypted pipe.

Those measures seem at least partially designed to avoid running afoul of state and federal privacy laws, which offer a confusing and often conflicting patchwork of regulations about what constitutes illegal eavesdropping and what doesn’t.

But even if Siri Recap isn’t actually storing full recordings or transcripts of overheard conversations, the mere act of “intercepting” them without notifying the involved parties could land a user in legal hot water.

“As I understand this Siri Recap feature, they don’t have any kind of notice or audible thing that’s happening, and so the individual who’s being recorded in this case is really provided no notice,” said Brian J. McGuiness, a partner at Indianapolis law firm Barnes & Thornburg and an expert in privacy and AI issues.

“So on the one hand, [Apple is] putting in some pretty serious privacy-by-design things in that they don’t necessarily have to and others aren’t doing,” McGuinnis continued. “On the other hand, they haven’t really gone the full way to say, we really want people to know when this device is being used or this feature is being used. [Apple] kind of left it up to the user … whether or not you want to tell people this is happening.”

In a support document, Apple says you should “be mindful of those around you” when using its new Audio Intelligence features, including Siri Recap. “While audio is not recorded and stored, and no speakers are attributed, consider those around you where conversations might be private or sensitive.” I’ve reached out to Apple for furthen comment.

Aside from the legal issues surrounding Siri Recap, there’s also the ethics and overall “ick” factor of microphone-equipped AI wearables in general, which may be listening at any given moment.

“Our right to conversational privacy must include freedom from other people, without our clear opt-in consent, using technology to document what we are saying,” Electronic Frontier Foundation Privacy Litigation Director Adam Schwartz said in a statement to PCWorld, adding that the organization is EFF is “disappointed” Apple is planning to ship the Siri Recap feature. “We suggest people should think twice before using this technology, out of respect to the conversational privacy of others.”

Of course, the Apple Watch isn’t the first AI-enabled device that can listen in on conversations. From Otter, the popular AI transcription app that can attend Zoom meetings (and is now embroiled in a privacy lawsuit) to Meta Glasses, AI gadgets and wearables with built-in microphones are becoming increasingly commonplace. Siri Recap on the Apple Watch appears poised to accelerate the trend.

“Thinking back, you know, you go back to the old … ‘is this creepy’ meter, and everybody would think that carrying around an always-on microphone, hidden in a bag or a hidden camera in a buttonhole, is creepy,” Barnes & Thornberg’s Brian McGuiness told me. “But as we normalize this and make it so prevalent and accessible to the average consumer, you’re only going to see more and more people getting more and more comfortable with the idea of recording these things.”

Prompt of the week: The “give outcomes to your agents” prompt

Unlike typical AI chatbots, AI agents like Gemini Spark, Claude Cowork, ChatGPT Work, and (now) Meta Muse don’t need prompts so much as they need outcomes. Instead of telling them how to behave, you want to tell your agents what you want at the end, whether it’s the lowest prices for holiday airline tickets or getting that pothole in front of your driveway fixed.

Using OpenAI’s new GPT-6 “Astra” model, I’ve crafted a prompt that can help you create a fleshed-out agentic outcome, complete with guidelines for keeping your AI agents on track, a clear finish line for the task, and instructions on when it should reach out to you for help.