Back in February 2012, w0rmer was on a hacking spree.
From Texas's Department of Public Safety, to Alabama's Department of Public Safety, to a police chiefs' association in West Virginia, to a county government in Alabama - he'd get in, grab whatever confidential and personal information he could find, erase data, and then deface the sites on his way out. He ran with CabinCr3w, a hacker group that had spun out of Anonymous. In his day job he was a Linux Systems Administrator.
By all measures this was a highly tech-literate individual, who wielded technology as a weapon, and was careful about everything a careful person is supposed to be careful about.
But then he broke into an FBI criminal database, and on his way out he left his calling card - a photo of a woman in a bikini, holding up a handwritten sign that read "PwNd by w0rmer & CabinCr3w, <3 u BiTch's!" Yes, very 2012.
This one time w0rmer had forgotten to strip the photo of its metadata. The woman who had taken that photo had done so using her iPhone with location services on. Through this, the FBI traced the photo back to her exact coordinates, discovered her identity, and found her Facebook page, which revealed w0rmer as her fiancé, Higinio Ochoa. He was arrested in Galveston, Texas, on March 20, 2012.
One photo got one man caught this one time because of one mistake. Now imagine if the FBI had 10,000 photos to work with.
If a picture tells a thousand words,
what do 10,000 pictures say about a person's life? Go ahead and upload any photo, say the last one you took, on https://theyseeyourphotos.com. You will get an immediate output based on what the machine sees in it - who's in the frame, what they're doing, what's in the background, and a few guesses about the people themselves. If you find it a bit unsettling, remind yourself that it's all based on just one photo. What happens when you add the other 9,999?
The first thing that appears is a map.
Every photo drops a pin, and after 10,000 photos, the pins huddle together in a handful of places. The one you're at every night and every weekend is your house. The one you're at on weekdays from 9-6 is your office. Fifty pins in a single weekend, and someone in a white dress in most of them, is where the wedding was. The one that shows up at 8am and 3pm during the week, with a kid in the frame, is the school. The one you go to once a year, and features older people, is where the grandparents live. The restaurant that is pinged again and again is the favourite of your family. You get the picture.
The second thing apparent is a network.
Your photo app already groups every face that exists across your library, so it knows who's in your life and roughly how much - the people in thousands of photos versus the ones in fifty. Based on this frequency, it is possible to infer who the most important people are in your life. Put that on the map, and suddenly the relationships are also clarified - the partner, the children, the colleagues, the friends.
The third, and maybe the most gut-wrenching realization, is the story.
You see, your photos always have faces. But their presence and absence, over time, tells your life story. A new face arrives in 2017, and is in everything by 2018. A dog arrives in 2019. A baby shows up in 2020. One of the original faces disappears in 2022, followed by 50 pins in one location, with everyone dressed the same and no one smiling.
This is the distillation of your life, indexed by who, where, and when, kept on a server that isn't yours. Ten years of friendships, relationships and breakups, written up for an algorithm to process. For maybe as mundane an output as 'What range of phones should we serve to this person in ads?'
Google says that Photos data isn't used for ads.
There's no real reason to not believe them. But Google Photos has the readable version of all your photos. The AI already scans them 'to improve our services'. All that stands between today and a full report on the last ten years of your life is a policy decision. And if that happened to change tomorrow, all you would get is an email about updated terms of service.
Take control.
With end-to-end encryption the sorting still happens - faces, places, trips - but it happens on your phone, and what reaches the server is something nobody can open. The decision to 'not look' isn't based on pinky promises any more, it's built into the architecture. Google says they won't look. At Ente, we couldn't if we tried.