Elon Musk’s X is the latest social network to launch a feature to label edited images as “manipulated media,” if posts by Elon Musk are to be believed. But the company has not explained how it makes this determination, or whether it includes images that have been modified using traditional tools, such as Adobe Photoshop.
So far, the only details on the new features come from a mysterious X post from Elon Musk said, “Revised visual warning,” as he re-shown the announcement of the new X feature made by anonymous Account X DogeDesigner. The account is often used as a proxy for introducing new X features, as Musk will repost it to share news.
Still, details about the new system are thin. DogeDesigner’s post claims X’s new feature could make it “harder for legacy media groups to spread misleading clips or images.” It also claims the feature is new to X.
Before it was bought and transformed into X, the company was known as Twitter has labeled tweets using media that is manipulated, fraudulently altered, or created as an alternative to omission. The policy is not limited to AI, but includes “selective editing or trimming or slowing down or overdubbing, or manipulation of subtitles,” said the site’s head of integrity, Yoel Roth, in 2020.
It is not clear if X uses the same rules or has made significant changes to overcome the AI. Its help documentation now he said there is a policy against showing inauthentic media, but it is rarely implemented, as the new deepfake debacle users share non-consensual mute images. In addition, even the White House now shows manipulated images.
Calling something “media manipulation” or “AI image” can be nuanced.
Because X is the playground for politics propagandaat home and abroad, some understanding of how companies determine what is “edited”, or perhaps AI-generated or AI-manipulated, should be documented. In addition, users should be aware that there is any dispute process beyond X’s crowdsourced Community Notes.
Techcrunch event
San Francisco
|
13-15 October 2026
As Meta discovered when it introduced AI image labeling in 2024, detection systems are prone to mistakes. In that case, The meta was found to be incorrectly tagged real photos labeled “Made with AI”, even if they haven’t been made using generative AI.
This is happening because AI features are increasing which is integrated becoming a creative tool used by photographers and graphic artists. (The new Apple Creator Studio suite, launched todayas a recent example.)
As it turns out, this Meta identification tool is confusing. For example, Adobe’s cropping tool flattens images before saving them as JPEGs, triggering AI Meta detectors. In another example, Adobe’s Generative AI content, which is used to remove objects — like wrinkles on a shirt, or unwanted shadows — also causes images to be labeled “Made with AI,” when they’ve only been edited with AI tools.
finally, Meta updated the label to say “AI Info,“To avoid immediately labeling an image as “Made with AI” when it isn’t.
Currently, there is a body setting standards for the verification of authenticity and proof of content for digital content, known as C2PA (Coalition for Content Provenance and Authenticity). There are also related initiatives like WATERor Content Authenticity Initiative, and Project Originfocus on adding tamper-proof provenance metadata to media content.
Of course, the X implementation will be subject to some process known to identify AI content, but the owner of X, Elon Musk, did not say what it was. He also doesn’t clarify if he’s talking specifically about AI images, or just anything that isn’t a photo uploaded to X directly from your smartphone camera. It’s not even clear if the feature is new, as DogeDesigner claims.
X is not the only outlet grappling with manipulated media. In addition to Meta, TikTok has also labeled AI content. Streaming services like Deezer and Spotify also scale initiatives to identify and label AI music, as well. Google Photos uses C2PA to show how photos on the platform are made. Microsoft, BBC, Adobe, Arm, Intel, Sony, OpenAI, and others are active C2PA steering committeewhile many other companies have joined as members.
X is currently not listed among them memberthough we have reached out to C2PA to see what has recently changed. X usually doesn’t respond to requests for comment, but we’re still asking.