You unlock your phone and there’s a notification: “Your memories from this day.” Sometimes it’s a nice surprise — a trip, a birthday, a pet. Sometimes it’s a photo of someone you’ve lost, a place tied to something painful, or a moment you’d rather not have resurfaced without warning. The feature that’s supposed to delight you doesn’t have a way to know the difference.
In March 2026, this stopped being a hypothetical concern. A man in Memphis filed a federal complaint after a photo of his recently deceased mother appeared as an automated “memory” on her own phone, days after her death — surfaced by the same algorithmic feature that, for most users on most days, just shows a nice throwback photo. The case is extreme, but it’s an unusually clear illustration of a feature that runs on everyone’s phone, working exactly as designed, with no concept of context.
How “Memories” Features Actually Work
Algorithmic memory features — found in Google Photos, Apple Photos, and most major photo apps — work by combining a few signals: the date a photo was taken (especially “on this day” anniversaries), facial recognition to group photos of the same people, location data, and sometimes broader categorization (trips, pets, events).
The system then assembles these into a “memory” — a curated mini-collection or highlight reel — and surfaces it through a notification or a prominent spot in the app, often without you asking for it.
This is genuinely useful when it works well. The problem is that the inputs (date, face, location) are observable facts, but the thing that determines whether a memory is welcome — the emotional context around it — is not observable to the algorithm at all. A photo from exactly one year ago is, to the system, just a photo from exactly one year ago. Whether that date now represents an anniversary worth celebrating or a date you’d rather not be reminded of is information the algorithm doesn’t have and structurally can’t have.
Where This Goes Wrong
Photos of people who have died. Facial recognition continues to group and resurface photos of someone after they’ve passed away, because the system has no way to know that fact unless explicitly told — and even then, “told” usually means a setting buried several menus deep, set after the fact, often after a painful surprise has already happened.
Photos with an ex, or from a relationship that ended badly. “On this day” features regularly surface anniversaries of trips, dates, or events with people you’re no longer in contact with — sometimes years after the relationship ended, because the photo’s date hasn’t changed even though its meaning has.
Photos taken at sensitive locations. Researchers studying Apple’s Memories feature have specifically examined the question of “sensitive locations” — places like hospitals, addiction treatment centers, or sites of personal trauma — and how (or whether) automated memory systems should avoid surfacing photos taken there. The fact that this is an active area of research indicates the problem isn’t fully solved by any major platform.
Photos that were never meant to be “memories” at all. Screenshots, documents, photos taken for a single practical purpose — these can get swept into date- or face-based groupings and surfaced as if they were meaningful keepsakes, simply because they happen to match the criteria the algorithm uses.
The Platforms Are Responding — Slowly
To their credit, major platforms have started adding more controls. Google Photos added settings giving users more control over who shows up in Memories — letting people remove specific individuals from being featured, which directly addresses the “deceased person” and “ex” scenarios. Apple has mechanisms to exclude certain people or photos from Memories and has done research specifically on location sensitivity.
But these controls share a common limitation: they’re almost all reactive. You typically discover the need for a setting only after a memory has already surfaced and caught you off guard — at which point you can prevent it from happening again, but the moment itself has already happened. There’s no way to proactively flag “this entire category of photos should never be resurfaced as a happy memory” before the algorithm has had a chance to try.
What You Can Do Now
Audit “memories” or “for you” settings on your main photo app. Look specifically for options to exclude people, exclude date ranges, or exclude locations from automated highlights. If you’ve experienced a loss, a breakup, or another life change in the past year, this is worth doing proactively rather than waiting for a notification to prompt it.
Turn off notifications for memory features, even if you keep the feature itself. A lot of the harm comes from the surprise — a memory appearing in a notification at an unexpected moment. Keeping the underlying feature but removing the push notification means you encounter memories only when you deliberately open that part of the app, which gives you more control over your own state of mind in that moment.
Be aware that “delete the photo” and “stop it being a memory” aren’t always the same action. Some platforms cache or pre-generate memory collections, meaning a memory already assembled might still surface even after you’ve deleted or hidden the source photo, until the next refresh cycle.
Separate “everyday photo library” from “personal archive.” Photos that are emotionally loaded — of people you’ve lost, places tied to difficult periods, documents and screenshots — don’t need to live in the same library that’s actively being mined for date-based and face-based “memories.” Keeping them in a separate, deliberately organized space means they’re preserved (you haven’t deleted anything meaningful) without being subject to an algorithm’s idea of when you’d like to be reminded of them.
Where daftei Fits
daftei doesn’t run facial recognition or algorithmic “memories” features that surface photos to you unprompted based on dates or detected faces. Organization in daftei is driven by you — through journeys and albums you create — rather than by an algorithm deciding what counts as a meaningful moment and when to remind you of it.
This isn’t a criticism of memory features in principle — for many people, most of the time, they’re a pleasant feature that works as intended. But “most of the time, for most people” is exactly the design assumption that fails the people for whom a particular photo, on a particular day, isn’t a pleasant surprise. A personal archive that doesn’t make that assumption — where what surfaces and when is something you control — is a meaningfully different relationship with your own photos.
Files in daftei are encrypted in transit with TLS 1.3 and at rest with AES-256, daftei doesn’t run facial recognition or AI modeling on your content, and doesn’t use what you store to train any AI system. For the photos and files that carry the most weight — the ones where an unexpected resurfacing would actually matter — that’s a deliberate design difference, not just a privacy one.
The Underlying Point
Algorithmic memory features are a good example of a broader pattern: software built around what’s technically possible (group photos by face and date, surface them periodically) rather than what’s emotionally appropriate (know when a photo represents something someone is ready to revisit, versus something they’re not).
The Memphis case is the most visible recent example, but the underlying mismatch exists every time these features run, for everyone — it’s just that most of the time, the photo that gets surfaced happens to be a welcome one, so nobody notices the system had no idea either way. As more of life gets photographed and more of those photos get run through automated “memory” systems, the gap between what an algorithm can detect and what a person actually wants reminded of isn’t going away on its own. The most reliable fix, for now, is keeping the photos that matter most somewhere that doesn’t try to guess.