➡ The author warns about the dangers of AI embedded devices like iPhone 16 and Windows Co-Pilot, stating they compromise end-to-end encryption and privacy. The AI in these devices can scan and remember all activities, making encrypted apps useless. The author traces this issue back to the San Bernardino terrorist attack in 2015, when the FBI wanted Apple to decrypt an iPhone, which Apple refused. Since then, technology has evolved to scan content before encryption, leading to a loss of privacy.
➡ AI technology in devices like the iPhone 16 and Windows Copilot PC can now monitor and analyze everything you do, including what you type and say. This technology can potentially report back to the company headquarters, raising privacy concerns. However, companies like Apple and Microsoft claim that data from the AI will never leave your device. To ensure privacy, it’s recommended to use devices without embedded AI companions, like a Linux PC or a de-google phone.

Have I been talking to deaf ears all this time? Apparently so. While my demonetized iPhone 16 video already exceeds 1 million views, apparently many are still buying iPhone 16. And I’m sure many are also getting excited by the new user interface of Windows Co-Pilot. So, my videos get watched but are ignored anyway. But I have to tell you something specific now because some of you are just not getting it. If you’re using these devices that have embedded AI like Apple Intelligence or Windows Co-Pilot, then don’t talk to me about end-to-end encryption. Because for you, it is officially dead.

We now have to rethink our communication methods, and I’ve talked about this since last year. The threat began with the technology called client-side scanning, and that has evolved. Now, this has expanded to AI-managed client-side scanning. This is the new world, folks, and even for those of us who don’t use these dangerous devices, it will now be important to know who you’re talking to and what device they use, because without your knowledge, you can get compromised. I know this kind of topic brings out the, I have nothing to hide people. You know what? If you know these people with nothing to hide, they are now a liability to the rest of us.

They, the normies, are the ones that will kill end-to-end encryption. They will be the people that will guarantee that the majority of the world will be under mass surveillance. It will be pretty hard to escape. If you want to understand what this threat is all about and how not to think like a normie, stay right there. I’ve replayed this little clip from Microsoft over and over, and I may have to keep replaying this so you understand what is being said. We are on a mission to create a true AI companion. And to me, an AI companion is one that can hear what you hear, and see what you see, and live life essentially alongside you.

Mustafa Suleiman in this video is the CEO of Microsoft AI. Let me re-emphasize what is being said. Your AI companion will see what you see, hear what you hear, know what you know. I don’t know why this is so hard to understand. So let’s translate this again a different way. You will have no secrets from the AI, zero. A friend of mine actually watched my iPhone 16 video and said, I will not buy an iPhone 16. Then I saw him over the holidays with a brand new iPhone 16. Then he contacts me in signal and reveals something to me that is utterly private.

Something that should have only been said face to face, and this is signal from his iPhone 16. So I had to remind him, did you know that you just told Apple that? And he replies, yikes, and then proceeded to delete the signal message, like that’s going to do something. Using any end-to-end encrypted app is useless when the spyware is watching the sucking screen, and it has a memory that’s sharper than an elephant. And no, deleting doesn’t do anything. See what you see. Watching a video, hear what you hear. Anything you do on the device is known.

Know what you know. Recently Zuckerberg was interviewed by Joe Rogan and the discussion apparently went to the end-to-end encryption capability of WhatsApp. What was interesting was that Zuckerberg acknowledged that the CIA can override your end-to-end encryption and can read your content anyway. That’s what he said. And of course, the security expert said ex-community notes chimed in and said that this Zuck statement is misleading because they said he meant that the ability to read your encrypted messages is from hacks to the phone like Pegasus. Apparently the community knows people think they’re experts but don’t understand the problem.

Sure, it is possible that someone hacked with a Pegasus malware could be compromised with a Trojan horse software that can read the screen. However, this is such a rare possibility since so far only a few enemies of some state have been targeted by Pegasus malware. This kind of attack is a state-level hack that results in millions in payments to the creator of Pegasus. It is not a mass surveillance tool. The bigger risk is the more obvious one, the AI that knows you. Is this such a difficult concept? The AI companion knows your every thought and your every interaction on your device.

It also understands the context of what you’re doing and is scanning your activities. The AI watches your screen so it doesn’t need to intercept each app. The AI watches what you type, the AI can do voice recognition, the AI can do text recognition, the AI can utilize other sensors on the phone. How are you going to hide what you’re doing? How is Signal or WhatsApp supposed to encrypt when the AI is scanning your screen before you even hit send? Let me repeat this history here so you understand the evolution of this.

You might think this is just a random fluke of technology that brought us here today. Granted, the fast development of AI is now part of this, but let’s follow the sequence. This started with the San Bernardino terrorist attack in 2015, now close to 10 years ago. Those terrorists were killed in a shootout so they will forever be silent. But they left their iPhone, so the FBI filed an application for a warrant to require Apple to assist in breaking into and decrypting the contents of that iPhone. This became a big public dispute because Apple refused to do that.

Now many people like Bill Gates started shaming Apple into complying, but Apple stuck to its guns and refused. In retrospect, it is clear that if Apple succumbed to the social pressure and aided in the decryption of that phone that today they would not be able to sell the iPhone as some sort of privacy device, which they started to do five years or so later. They would not be the largest corporation in the world. They came up with the mantra, iPhone is privacy. They still claim this today, but have you noticed the silence from the three-letter agencies? How come no one is on the news attacking the latest iPhone due to its unbreakable encryption? Let me tell you that CIA Director Brennan at the time was constantly on the news griping about end-to-end encryption.

And in the last few years, there’s been nothing but crickets. One thing you learn in intelligence gathering is that you watch the signals. And an absence of signals mean as much as the preponderance of signals. It should be obvious that the intelligence agencies are now appeased. And just like Zuckerberg says, three-letter agencies can break into an encryption, but apparently the average person doesn’t know how it’s done. So the first progression to watch here is that immediately after the Apple FBI dispute, the three-letter agencies began to promote a solution to the public.

They said the solution to breaking end-to-end encryption was to read the content pre-encryption. So in case you don’t understand what this means, it translates to capturing content locally on device and reporting what’s in that content before it is sent through whatever encrypted app. A few years ago, I think three years ago now, Apple announced the solution which we all now refer to as client-side scanning. And this is almost a direct response to what CIA Director Brennan was asking for. The original communicated solution was that your photo content would be scanned on your local device and then based on criteria provided by Apple, it could see if there’s some content there that’s deemed illegal.

So watch the flow here. Apple sends a list of things to scan for to every iPhone in existence and using image AI it would then report if the search criteria matches. If it matches, then the phone would be reported to Apple HQ and at the time they said they would forward that information to law enforcement. The stated purpose of this technology was to scan for illegal photos of children called CSAM. But this reveals some facts that you have to know. First, Apple can give instructions to your phones to scan for any particular content. The fact that they used the CSAM example does not preclude it from scanning your photos for, let’s say, general nudity or even political content.

All it means is that a channel was built in for the phones to report to HQ. This channel supposedly is encrypted and so we’re not able to observe the traffic. Again, from history, we now know that this client-sized scanning module is called Media Analysis D and is running both on macOS and iOS. So this is some sort of image processing AI, sort of a low-level AI that is not an LLM. Apple stated, by the way, that they suspended the CSAM project, but anyone with macOS can easily see that Media Analysis D continues to run and cannot be stopped.

So even a few years back, it is clear that the client-sized scanning was already built into iOS, though it seemed to only focus on image scanning. Now enter Microsoft and the new Windows Copilot. The clue to the current trend in technology became obvious when the feature called Windows Recall was introduced in Windows 11 in 2024. Windows Recall is actually a combination of several pieces of software. One is the screenshot capability. Every few seconds, a screenshot of your display is taken and analyzed, and the analysis is stored as text. Then this was also combined with key logging, so that Windows can see everything you type.

Then that is also stored. These two technologies, combined with other things like voice recognition, now will summarize in that new AI mantra, see what you see, hear what you hear, know what you know. So while the old photo client-sized scanning on Apple devices simply examined images in your file system, the new version of this now involves AI. Let me remind you again that Bill Gates was absolutely in support of being able to break encryption on devices, so this would be consistent with Microsoft’s objectives, and Apple’s now as well. AI can now observe everything you do, and because it no longer runs at the app level, it is able to intercept your traffic through the same interfaces you use.

These are the screen, the keyboard, the speakers, and the microphone. Now we have an AI fully aware of its environment. Just like Windows Copilot, Apple Intelligence was announced as being integrated into the iPhone 16. Now both these operating systems, Windows and iOS, have limitations here. In order to do this kind of AI spyware, the CPU and the NPU have to be powerful enough to not interfere with the speed of your interaction with the device. This is why Windows Copilot is intended to be implemented only on Copilot PC branded computers, and Apple Intelligence is intended for iPhone 16s and M-Silicon Mac OS PCs.

Will AI talk to HQ? Understand the little clues I left for you here to figure it out. Apple already demonstrated with their own words that they could control client-side scanning remotely. Now Apple and Microsoft made zero statements about the possibility of the AI being able to report to HQ, and let me tell you that technically speaking, this is a trivial task. The way it would work would be that HQ would send an instruction to all devices with AI searching for a particular scenario, and then the device AI that finds a match will raise its hands to HQ.

I’ll give you an example that China would love to ask the AI. Find people in China that have mentioned or are discussing the Tiananmen Square Massacre. In China, it is illegal to discuss this as it is something they intend to erase from history. So those wanting to discuss this will not have to go to an isolated countryside with no devices because their own phones can now spy on them. Let me tell you some truths that Apple and Microsoft will say, that data from the AI will never leave your device. I think this is correct.

If, for example, there are nude photos on your phone, it is expected that only the AI can see it and that I don’t expect your photos to be forwarded to Apple, for example. But it should be obvious that Apple can ask which phones have nude photos and more precisely, what other specific things are found in those photos beyond just nudity. How about describing the age and gender of the parties, their facial expressions, and sounds heard? Are there guns? Are there weapons? And it should be obvious, too, that this technology has now evolved.

The iPhone 16, just like Windows, has a device that’s always listening. The microphone is always on, the camera is always on, the infrared sensor is always on, the location is always on. So beyond just scanning photos, the nude tech means your AI companion is practically alive. Any wrong move and wrong thought from you will potentially trigger the thought police. What can HQ ask the AI? Obviously, the AI can read passwords, see your seed in your crypto wallet, see your bank accounts. There is absolutely nothing that can be secret if you do it on the device or within the sensing capability of the device.

And this is what you want. You’re basically buying a tool of enslavement with your own money. And supposedly because you can now take better pictures on your sucking iPhone, and Windows will solve your problem of forgetfulness. I said this a long time ago. The first step that gives us a clue of the danger of this AI technology is when the AI becomes your friend. And here we are. The practical effect of this technology is that presumably the bulk of the population will be using a tainted device. In a few years, I expect the majority of people will have an AI companion device.

We know that end-to-end encryption still works assuming there is no spyware watching your screen or capturing your keystrokes. But now it is not sufficient to know that your device itself is safe. Everyone participating in an encrypted conversation needs to also be on a safe device. A device with no embedded AI companion. For example, this would be a Linux PC or a de-google phone. You can see the problem here. The only safe apps are ones that will detect if people you are conversing with are on a safe device. And if not, then you should assume your conversation is not a secret.

I will pass this tip on to makers of encrypted apps. For example, I can pass this tip on to session developers. In the meantime, tell your friends that if they want safe conversations they cannot be on an iPhone 16 nor use Windows co-pilot on a co-pilot PC nor use macOS on an mSilicon Mac. Eventually, we likely have to add Google Gemini to this list using the Pixel 9 Pro. Temporarily, old iPhones will still be passable and old Google androids. The Google font and Linux will continue to be safe but now we have to be aware of the other party we are communicating with.

Folks, this channel relies on your community support. As you’ve seen, I don’t take any sponsors and instead of asking for donations we have created solutions to aid in the privacy battle ourselves. One of the newest solutions is the Brax 3 phone which is currently on the crowdfunding site indiegogo.com. This phone is a true privacy phone which we created with a team of contributors. Check it out. We also have other products like the Brax Virtual Phone which is a voice over IP service with no KYC. A rare feature nowadays. We also have Brax Mail that allows for unlimited aliases so you can use a different email for every internet platform.

We have Byce VPN which is our reasonably priced VPN solution. We also offer other de-google fonts which will provide the benefits I’ve discussed in this video. All these are on my platform, Brax Me. Please join the community of over 100,000 users there talking about privacy daily. The store is there with the products I mentioned. Again, your support sustains this channel and thank you for that and your donations to Patreon, Locals and YouTube memberships. Thank you for watching and see you next time. Thank you. [tr:trw].

See more of Rob Braxman Tech on their Public Channel and the MPN Rob Braxman Tech channel.