What I discovered about camera-roll AI, algorithmic profiling, and taking back control of my devices
I am not on social media very much anymore and that has changed my experience of these platforms in ways I did not expect.
Recently, I received an email from one of the platforms essentially telling me that because I was not participating enough, changes were being made to how my profile and my Kanien’kéha (Mohawk language) experiences that I had created on the platform were being handled.
That got my attention.
Then there was the feed.
I started seeing increasingly disturbing AI-generated or AI-styled cartoon characters that looked remarkably like me and my husband: similar hair, facial characteristics, body shapes, our specific clothing and overall appearance.
But there was something about this that I found particularly disturbing.
I did not have photographs or video of my recent hairstyles, clothing choices or faces in my camera roll.
Yet I was being presented with cartoon characters that appeared to be designed to look like what I was wearing that day, including my hairstyle and, yes, my face, from different angles.
You can imagine how disturbing that was and that the experience raised a very serious question for me:
How did the system generate such specific representations of how my husband and I looked at that particular time?
The stories accompanying some of these images were even stranger.
They repeatedly pushed extremely negative themes about aging, body image, marriage, appearance, clothing choices and personal worth.
According to the messages being presented to me, my wrinkles were apparently offensive to everyone around me.
My body was supposedly embarrassing.
My clothing choices were unacceptable.
Even my hair, which I happen to like very much, was apparently something everyone else was disgusted by.
Then the recommendations expanded beyond my appearance.
The brand names of my clothing, household cleaning supplies, shampoo, food and water apparently needed to be completely replaced with what the social-media platform had decided were better brands.
Many of the recommended products appeared to have essentially the same ingredients as products I already use, except that most of the recommendations from the cartoon character versions of me were insisting I change to their brands which are being shipped from China while I live in North America.
At some point I had to step back and ask:
What exactly is this system doing?
And perhaps more importantly:
How are ordinary human beings, with no training in protecting themselves from algorithmic manipulation, gaslighting, emotional abuse and psychologically targeted content, supposed to respond to an experience like this?
I found the whole thing profoundly bizarre and disturbing.
And I want to be very clear about something:
I do not accept an algorithmically generated feed as an authority on my life, my body, my marriage, my culture or my worth.
But the experience did make me ask a much more technical question:
What exactly are these platforms allowed to see?
1. Your camera roll may be more connected to Social Media than you realize
This part is not speculation.
Meta has introduced a Facebook feature called camera-roll sharing suggestions.
According to Meta, when the feature is enabled, Facebook can analyze photos and videos in your camera roll and use AI to suggest things such as edits, collages and videos.
Meta describes this as an opt-in feature.
That distinction matters.
This is not the same thing as saying:
“Facebook secretly steals every photo on your phone.”
The documented feature is more specific:
If you give Facebook permission to use the camera-roll feature, your previously unposted photos and videos can become inputs to Meta’s cloud-based processing system.
Meta says the suggestions are private to you and that you decide whether to share them.
Meta also says camera-roll photos and videos are not used to improve its AI unless you choose to publish or share them through an AI interaction.
But there is still a basic privacy question worth asking:
Why should an app need access to photographs I have never chosen to upload?
For me, that question is enough reason to turn the feature off.
2. Check whether camera-roll processing is enabled
The exact wording and location of settings can change as Facebook updates its apps, so don’t assume your screen will look exactly like mine.
Look in Facebook’s settings for the camera-roll sharing/suggestions controls.
The setting I was looking for was:
Camera roll sharing suggestions
and specifically the option associated with:
“Get creative ideas by allowing camera roll cloud processing.”
If it is enabled, turn it off if you don’t want Facebook processing your camera-roll media.
Meta says the feature can be disabled at any time.
3. Then lock it down at the operating-system level
This is the step I consider more important.
Don’t rely entirely on an app’s internal privacy switch.
Go into your phone’s operating-system permissions and review what Facebook, Instagram, Threads, Messenger and other social-media applications can access.
On iPhone
Go to:
Settings → Privacy & Security
Then review:
- Photos
- Camera
- Microphone
- Contacts
- Location Services
- Bluetooth
- Local Network
For Photos, consider using:
None
or
Selected Photos
rather than giving an application unrestricted access to your entire photo library.
On Android
Go to:
Settings → Apps → [app name] → Permissions
Review:
- Photos and videos
- Camera
- Microphone
- Location
- Nearby devices
- Contacts
Set permissions to the minimum level you actually need.
The principle is simple:
If an application doesn’t need access, don’t give it access.
4. Don’t forget Instagram and Threads
Facebook isn’t the only Meta application on your devices.
Review the permissions for:
- Threads
- Messenger
- Meta AI
- Any other Meta application you have installed
Meta has been rapidly integrating AI capabilities across its products, including image generation, image editing, photo understanding and AI-assisted experiences.
That makes understanding your permissions more important, not less.
5. AI can recognize far more than a face
This is another reason I pay attention to what I allow these systems to process.
Modern computer-vision systems don’t simply answer:
“Is there a face in this picture?”
They can analyze visual characteristics, objects, clothing, environments, poses, micro expressions, relationships between objects and many other attributes.
Meta itself describes its AI image systems as using computer vision and machine learning to understand what is in an image and apply transformations to it.
And Meta’s newer image-generation systems can combine multiple photographs and other visual references to create new images.
So I don’t think of a photograph merely as:
a photograph.
I think of it as:
a machine-readable bundle of information about the people, places, objects, clothing, environments and circumstances represented in that photograph.
That distinction matters.
6. What about the microphone and camera?
This is where I want to be careful.
I have seen plenty of people claim that social-media applications are secretly turning on microphones and cameras and recording everything around them.
I’m not going to tell you that strange advertisements or strange AI images prove that your microphone was secretly recording you.
They don’t prove that.
But I do believe in minimizing unnecessary access.
If Facebook doesn’t need my microphone, I turn it off.
If Instagram doesn’t need my camera, I turn it off.
If an application doesn’t need access to my photographs, I turn that off too.
That’s not paranoia.
That’s basic permission management.
7. I put physical covers over my cameras
I use physical sliding covers over the cameras on my computers and other devices.
They’re inexpensive, removable and don’t require me to trust an application to behave correctly.
I noticed something interesting after I started doing this.
The strange cartoon character versions of me that had been appearing in my social-media experience stopped appearing.
I am not presenting that observation as proof that my camera had been secretly accessed.
There are other possible explanations.
The timing could be coincidence.
The images could have been generated from photographs I had previously uploaded.
They could have been generated from other information available to the recommendation system.
I simply don’t know.
What I do know is that the physical camera cover gives me one thing that a software permission screen cannot:
a physical barrier.
And I like that.
8. Physical microphone blockers are a different matter
There is another important technical distinction here.
A USB-C blocker, Lightning-port blocker or 3.5 mm accessory can prevent access through a particular physical port.
It does not disable the built-in microphone inside your phone or laptop.
So don’t buy a little plug and assume you have physically disabled every microphone.
If you want physical protection against an external microphone being connected through a port, a port blocker can be useful.
For an internal microphone, you need to use the device’s hardware controls, operating-system permissions or a physical microphone-disconnect solution designed for that particular device.
The same principle applies to cameras.
A physical camera shutter can block the lens.
That’s different from proving that software did or did not attempt to access the camera.
9. The disturbing part isn’t only surveillance
For me, there is another issue that is becoming much more important.
Algorithmic psychological profiling.
My concern isn’t simply:
“Can the platform see my photograph?”
It is:
“What does the platform infer about me from everything it can see, measure and observe?”
And then:
“What does it show me because of those inferences?”
That is a much bigger question.
A recommendation system doesn’t need to literally know that you are ashamed of your body in order to show you thousands of messages about body dissatisfaction.
It may simply learn that particular imagery keeps your attention.
The same applies to aging, relationships, appearance, insecurity, fear, outrage and anxiety.
An algorithm doesn’t have to hate you.
It doesn’t even have to understand you.
It only has to learn which signals keep you looking.
That is arguably more disturbing.
10. My body does not belong to the algorithm
One of the strangest consequences of this experience was realizing how aggressively some platforms can insert themselves into a person’s relationship with their own body.
Apparently I’m supposed to be embarrassed about aging.
Apparently I’m supposed to hate my wrinkles.
Apparently I’m supposed to believe my body is embarrassing other people.
Apparently I’m supposed to change my clothes, my hair and my appearance because an algorithm has decided those things are problems.
No.
My body is mine.
My marriage is mine.
My aging is mine.
My culture is mine.
And my relationship with Kanien’kéha is mine.
An algorithm does not get to define any of those things for me.
Ironically, I did purchase some compression leggings after being bombarded with body-related messaging, and I actually find them comfortable and useful.
That’s my decision.
There is a huge difference between:
“I chose something because I like it.”
and
“A machine repeatedly told me something was wrong with me until I changed myself.”
That distinction matters.
11. What I am doing now
I’m taking a much more restrictive approach to device permissions.
My basic rule is:
Camera
If I don’t need an application to use it:
OFF.
Microphone
If I don’t need an application to use it:
OFF.
Photos
If I don’t need an application to see my entire library:
OFF or Selected Photos.
Location
If the application doesn’t genuinely require location:
OFF.
Contacts
If the application doesn’t need my contacts:
OFF.
Background activity
If I don’t need an application constantly running:
RESTRICT IT.
Physical camera
When practical:
COVER IT.
12. We need a different relationship with AI
I am an applied AI engineer.
I work with artificial intelligence because I know it is extraordinarily powerful.
But powerful technology requires boundaries.
My philosophy is simple:
AI accelerates. Deterministic systems prove. Humans remain accountable. Systems remain governable.
That applies to social-media AI too.
We should not automatically accept every new AI capability simply because it is convenient.
“Allow” should actually mean:
I understand what I am allowing.
And “personalized” should not become a euphemism for:
a machine building an increasingly detailed model of me and then using that model to influence what I see.
13. Check your devices
If you have noticed unusually personal, disturbing or strangely specific content appearing in your social-media feeds, don’t immediately assume that your camera or microphone has been secretly activated.
Start with what you can verify.
Check your:
- Photo permissions
- Camera permissions
- Microphone permissions
- Location permissions
- Contacts permissions
- Bluetooth permissions
- Local-network permissions
- Background-app permissions
- Cloud photo integrations
- AI features
- Camera-roll processing settings
- Account privacy settings
Then reduce access to the minimum you actually need.
And if you don’t use a social-media application very much anymore, ask yourself a much simpler question:
Why is it still installed?
One final thought
I don’t want to live in a world where I have to wonder whether every photograph I take is becoming training material, every expression is becoming behavioral data, and every insecurity is becoming an engagement opportunity. I don’t even want to know what most random people are thinking or judging me with either because that is none of my business to be honest. This is basic 101 life lessons of being a decent human being and doing our best to be our best with our own individual life purpose and gifts.
I also don’t want an algorithm telling me what to think about my own face.
Or my body.
Or my marriage.
Or my age.
Or my culture.
Or my language.
Technology should serve human beings.
Human beings should not have to psychologically conform themselves to whatever the algorithm decides will keep them scrolling.
If you have had a similarly strange experience with Facebook, Instagram, Threads or another platform, I’d genuinely like to hear about it.
Not because every strange experience proves surveillance.
But because patterns are worth documenting.
And because the first step toward technological accountability is paying attention to what is actually happening.
Check your permissions. Reduce unnecessary access. Cover what you can physically cover. And don’t give an algorithm authority over your sense of self.
AIGeneration.blog 