Last week, Instagram automatically began labeling photos that had been altered with artificial intelligence tools as “Made with AI.” On the surface, this sounds like a victory for photographers who denounce the use of AI as intellectual property theft. But in reality, there’s little room for nuance or debate in the way Meta denounces AI.
Fasten your seat belt. This is probably a pretty extreme opinion, especially coming from a photojournalist.
Earlier this week, I used my phone to upload to Instagram a photo by renowned New York photographer Louis Mendes taken at Photoville. I did some standard Photoshop editing that’s common in my non-journalistic work, but in the process, I boxed off a small section at the edge of the frame and told the generative AI to remove the highlights. Here’s the before and after results:
Once uploaded, I was surprised to see that Instagram had labelled the post “Made with AI” at the top.
No, it’s not. You just lasso a small part of a photo and let the AI edit it for you. This is the same edit you would make with the clone tool and a little time. The end result will be nearly identical, except one photo will be flagged and the other won’t.
In this distinction, the broad phrase “Created with AI,” while noble in its pursuit of truth, fails. The label seems to imply that this image was created entirely with AI, which is not the case. The AI was used as a retouching tool, much like the dodge tool, burn tool, clone tool, healing brush, etc. To single out the use of generative AI tools in this label seems to misunderstand how AI was used in this case.
Of course, if DALL-E or Midjourney had generated this image of Louis Mendes out of thin air, the “created with AI” label would apply, but I don’t think that’s the case here, because Mr. Mendes was just standing there as I typed this sentence (the snakes and dinosaurs were not, but more on that later).
This can have a negative impact on retouching in general. Here’s another example where the “Made with AI” label makes no sense.
The only AI “crime” in this photo was that it was used to remove the front license plate of this car, which falls into the retouching category rather than making the image fully AI-generated. If edits like this are demonized, then why do AI tools exist in the first place?
There are other issues too. One of the commenters on my post asked if using Photoshop’s AI Denoise feature would trigger this label, which is very unflattering for event photographers who use the tool for their clients. I tried it and it seems Instagram doesn’t apply the label when you use that tool. Instagram’s help page has a cryptic explanation for the label, saying they look at “industry standard signals” to determine it. I’ve seen AI Denoise introduce nasty artifacts and fake faces into photos, and this isn’t immune to fakes either.
Additionally, the photo at the top of this post, despite being clearly created with AI and labeled as such, does not receive such a designation when uploaded from a desktop web browser. This is a pretty big loophole, and a pretty uneven application of Instagram’s AI policies.
Creatives have long been embracing new tools to make better art, whether for client work or for personal satisfaction, but it’s a huge mistake for a big company like Meta to appoint itself as the arbiter of how the “Made with AI” label should be applied — it stigmatizes those who are using AI responsibly.
Certainly, journalists and truth-tellers should not use such tools to edit their work. But should wedding photographers have to deal with lawsuits from brides who see the label “created with AI” when they upload their wedding photos to social media? Should companies that edit photos to hide wardrobe flaws on social media face backlash if the photos are labeled as AI?
These are questions that Meta doesn’t seem to have fully explored.
But it’s something photographers will definitely take into consideration.
What do you think about Instagram’s new “Made with AI” label? Let us know in the comments below.
