What comes online, goes AI
Trying to navigate the world of limitless surveillance, data accumulation, and AI training
Happy Friday, Control Spy, Delete readers! Time for our weekly update on what’s up with surveillance technology and online censorship in this world.
The deeper we get into the AI revolution unfolding around us, the more acute the question becomes: if I don’t want my data to become food for artificial intelligence algorithms, what should I do? If shutting down your social media accounts and hurtling your laptop into the sea is not an option, you might consider changing some of your online habits.
Guess what – I decided to ask AI itself. I asked Claude, ChatGPT and Gemini: “What are the best ways to protect my data from being used by AI algorithms like you?” And you know, they were helpful!
All three reasonably recommended to avoid putting any sensitive data online – as well as including it in your chatbot prompts. An example from me: if you’re asking ChatGPT about the pain in your stomach, maybe don’t tell it your entire medical history – or be prepared for that data to be used for commercial or other purposes in the future.
Then, the chatbots recommended paying attention to privacy settings and what you allow your apps to access – including whether you allow AI agents to use your dialogues for training. However, this recommendation comes with a big caveat all three admitted: there is no perfect protection against AI bots scraping and using your stuff.
“Remember that anything you post publicly may be collected by AI companies or web crawlers unless the site explicitly blocks them. No technical setting can fully compensate for sharing sensitive information publicly,” says ChatGPT.
“Regardless of the settings you toggle, assume anything you type into a public LLM could potentially be seen by a human reviewer,” says Gemini.
“Once something is public on the internet, it’s very hard to fully prevent it from ever appearing in a training dataset somewhere — scraping happens across many actors, not just one company,” says Claude.
So take it from these guys and hold onto your data if you can help it.
With that in mind, time to dive into this week’s stories!
Share this newsletter with your friends and followers if you’re enjoying it.
Biometrics briefing
A shooter who wounded one individual in DC on June 8 was identified using Clearview AI’s facial recognition tool. – DC Witness
The European Commission rejected a request by the airline industry to pause the new EU biometric border check system. – The New York Times
UK police have arrested 28 people in the city of Bradford since February using facial recognition technology. – BBC
Police in India are using gait analysis as evidence in a murder case. – NDTV
India’s Internet Freedom Foundation is asking the government to reconsider plans to introduce facial recognition at the country’s major airports. – IFF
Qatar’s Hamad International Airport is now allowing passengers to use facial biometrics instead of boarding passes to check in, drop off their bags, clear security, and board flights. – ID Tech Wire
Every move you make
Meta, the parent company of Instagram, Facebook and WhatsApp, has patented a system designed to record everything you say during the day to produce tailored workout recommendations. The patent, filed last year, was first reported by Patentlyze and then by 404 Media.
The patent describes a system that records human speech, transcribes it, and feeds it to an artificial intelligence algorithm, along with contextual data: when you were when you said it, what time of day it was, what were you doing, and what apps you were using at that moment.
AI also analyses the tone of voice and pace of speaking to capture the user’s emotions. Training data for such a product might also include users’ personal messages, books and newspapers they read.
“An implementation may show that the user laughs more often on certain days, shows improved mood after life events, or expresses more positive emotion during morning routines. The device may also provide citations to specific audio moments that support the emotional interpretation,” the document said, cited by 404 Media.
The idea of AI listening devices is not very new – a slew of AI note-takers in the shape of bracelets, pendants, or pins has already hit the market, offering to listen to and transcribe everything that happens during your day. Since the beginning of the AI hype, it’s been an open season for all human-generated data, which is vital for the development of the new generation of artificial intelligence algorithms.
But even in that context, capturing and analysing people’s emotions in real-time sounds sinister. Facebook has already experimented with analysing and impacting users’ emotions before: in 2012, the company enrolled about 700,000 users in a study of “emotional contagion,” altering their news feeds to see whether it would change users’ moods (yes, it did). The users were not warned.
The experiment showed what we now know pretty well: the content we consume impacts how we feel about the world and our own lives, which makes our social media feeds an extremely powerful manipulation tool. Now imagine that tool on steroids, armed with every bit of information about your private life and the capacity of the modern AI.
Remember the movie Don’t Look Up? There was also a device that read users’ emotions and used that data in real time in the story. The company behind the device was determined to capture and exploit everything it could touch, from human emotions to the literal end of the world.
But dystopian books and movies are considered manuals now, am I right? Next time someone writes a science fiction book – please, guys, be careful, you might be writing an actual scenario of our future downfall.
Waymo snitching on teens
But some of us are already living in the world of spying devices, and this time, I’m talking about self-driving cars. Such cars, to be functional and safe, have to be constantly monitoring their surroundings, as well as what’s happening inside them, which makes them perfect mobile surveillance devices.
And some California teenagers had a chance to experience that this week. On Monday, the San Mateo Police Department posted on Facebook that two 15-year-olds had been detained after they’d been rolling around in a Waymo car while drinking and shooting toy guns from the windows.
“Parents do you know where your teens are? @waymo does!” the post says, adding that the kids were “shooting Orbeez from the car as they sipped on afternoon libations while being chauffeured around town in the driverless vehicle.” The police also posted a video of several armed officers, one of them with a dog, approaching the stopped Waymo car.
The police spokesperson told NPR that the car was disabled before the cops arrived, so the teens could have left the scene but didn’t. They were ultimately released to their parents, but potential charges are still pending, she added.
And this case is actually interesting because if the passengers were armed with real weapons and posed an actual danger, I bet most of the people walking and driving on that street near that Waymo car would approve of the smart vehicle taking care of the situation before the police arrived.
But the matter of agency that we assign to the smart machines increasingly surrounding us is the trickiest part of this new surveillance age dilemma: do we still trust in humans’ ability to act right on their own, or would we rather surrender to the power of an artificial system that would be watching – and correcting – our behaviour for us?
Risk scores for basketball fans
Journalists keep revealing the details of a surveillance system inside Madison Square Garden (MSG), New York’s most famous arena and home to the Knicks basketball team. Last month, hackers breached MSG and published internal documents.
According to Wired, the company, lead by the owner James Dolan, has been keeping a dossier on MSG’s high-profile guests and assigning them “risk scores.” The database contains almost 40,000 entries, according to the report. A source inside the company told Wired that security personnel have monitored the social media accounts of celebrities who sought complimentary tickets to see if they post anything negative about MSG.
And by anything I mean anything – a New York City rapper Jadakiss, for example, was designated as “medium risk” because he posted that Dolan “seems like he’s always more happier when the team sucks,” according to Wired. Some celebrities are marked as “LGBTQIA.” There is also a list of famous people whom the firm decided not to provide free tickets to because they had been critical of the company’s management.
The MSG documents dumped online by hackers in June also contained personal data of millions of customers, including public officials like Jessica Tisch, commissioner of the New York Police Department.
The leak prompted a class-action lawsuit alleging that the loss of customers’ personal data was a direct result of James Dolan’s fixation on surveillance. MSG simply shouldn’t be collecting and storing all that personal information, Surveillance Technology Oversight Project legal director Darío Maestro argues.
“A company that cannot secure a customer list has no business scanning our faces,” he told Wired.
Watch the Watchers: Spyware victims strike back
Eight victims of Predator, the infamous spyware made by Intellexa, sued the Athens-based company in Greece, seeking 1 million euros each for moral harm, Reuters reported on Tuesday.
Back in 2022, multiple Greek politicians, civil servants, and journalists were targeted with Predator, causing a political scandal. As a result, the head of the state intelligence service and the prime minister’s chief of staff lost their jobs, and Intellexa was sued by the Greek government. This February, the firm’s top managers were sentenced to prison by a Greek court.
Another victim, Greek parliament member and former prime minister Antonis Samaras, filed a motion with the Prosecutor of the Supreme Court requesting a full investigation into how his phone was breached, To Vima reports. In May, the government blocked the creation of a new parliamentary committee that would investigate the spyware campaign.
Intellexa was sanctioned by the U.S. in 2024 for targeting U.S. citizens, among other victims, but in December 2025, the Trump administration lifted the sanctions on three people in Intellexa’s leadership, saying they had “demonstrated measures to separate themselves from” the firm.
Tips and Tricks: No Instagram for AI
The New York Times published an explainer on how to fence your Instagram profile off from Meta’s new AI image generator Muse. All adult users with public accounts are already automatically opted in. If you don’t want other people to use your pictures in their AI collages, go to the app’s settings, find “Share and reuse” and turn off the “Allow people to reuse your content on Instagram and with AI features” setting.
To be honest, my Instagram settings menu looks a bit different from what the NYT describes and doesn’t mention AI at all (I’ve updated the app), it only has an “Allow people to create with and reuse your content.” See what works for you.
The ultimate way to block any AI use of your pictures is to make your profile private – or rather, as we’ve learned from the AI agents themselves, not post thos pictures online at all.
***
And that’s a wrap for this week, guys.
Stay vigilant!
Anna