Instagram is mislabeling posts 'Made with AI'. What's in it for them?
We might be months away from 'Verified Real'
There’s a scene in Walt Disney’s 1951 classic Alice in Wonderland in which our heroine encounters an Ace, a Two, and a Three of Clubs, singing jauntily while haphazardly painting white roses with a thick coat of red. White ones were planted by the cards by mistake—by concealing them, they hope the error will go undetected.
It looks violent and messy—paint splattered everywhere, conveying blood dripping from Three’s throat as he gestures having his head cut off. The cards know the paint will suffocate and kill the delicate roses, yet they do it to please the Queen of Hearts. They’d rather keep their heads affixed to their bodies than upset her.
We're painting the roses red.
Oh, painting the roses red.
And many a tear be shed,
Because we know,
They'll cease to grow,
In fact, they'll soon be dead,
NOOOOO...
And yet we go ahead,
Painting the roses red.
In a similar haphazard fashion, Instagram is splattering some posts with a Made with AI watermark, seemingly to define and classify those images from conventional ones.
It makes sense for images obviously made with software like DALL-E, StableDiffusion, or Midjourney—I’m looking at you, Shrimp Jesus. But the problem arises when Instagram is mislabeling images because they contain only some elements of AI, such as Adobe Photoshop’s denoiser, or generative fill to remove a blemish.
Former White House photographer Pete Souza had an image falsely classified as Made with AI, because he used an AI-powered crop tool in Adobe Photoshop.
I’m not clear why Instagram is using the “made with AI” on my post. There is no AI with my photos.
With the prevalence of retouching tools enhanced with AI functionality in image editing software, it’s almost impossible to avoid AI interference, no matter how hard you try. Inevitably, there are varying degrees of it in most photos uploaded to Instagram. Meta didn’t create a sliding scale to indicate what percentage of an image is AI. Rather, they make a binary decision: this photo is AI, this photo is not.
Like the “Painting the Roses Red” scene in Alice in Wonderland, Meta’s Made with AI label is aggressive, intentionally deceptive, and self-preserving by inflating the number of AI images on its platform. I can only speculate on the reasons why…
Create the impression that Instagram is the dominant source for user-generated AI images, to handily prevail over competing but less mainstream platforms like Discord and Reddit
Generate a pool of images that are fair game for Meta’s AI scraping and research
Test public response to Meta’s overarching AI classification
Position the company as an authorized judiciary on what constitutes an AI-generated image from what is not: namely, what is real vs. what is fake
We may be only a few months away from a Meta verification stamp that defines whether an image is real or fake. That would make its opaque credential system the de facto authority on trustworthy images, for which this labeling is a precursor.
The Queen of Hearts noticed the disguised white roses because the sloppy paint job and sentenced the cards to death with a hearty “Off with their heads!” It is yet to be determined whether users will notice the wet paint and similarly condemn Meta.
