Meta attaches ‘Made with AI’ labels to real photos on Instagram, sparks outcry

In one instance, an Instagram user shared how a photo of their grandparents carried the AI label whereas another user said that an actual photo of herself in a costume had the controversial tag. Closer home, several posts by the official Kolkata Knight Riders handle have been marked with ‘Made with AI’, including a recent photo of the cricket team winning the 2024 Indian Premier League (IPL).

“We are taking into account recent feedback and continue to evaluate our approach so that our labels reflect the amount of AI used in an image. We rely on industry standard indicators that other companies include in content from their tools, so we’re actively working with these companies to improve the process so our labeling approach matches our intent,” a Meta spokesperson told indianexpress.com. Meta did not directly answer if it plans to temporarily pause the AI labelling feature on its platforms.

Ahead of the Lok Sabha elections in April, Meta announced that it would label AI-generated content and manipulated media on Facebook, Instagram, and Threads. However, the initiative by the Mark Zuckerberg-led company still has a long way to go. The posts reportedly mislabelled as AI-generated appear as such only in mobile view. Not just that, the AI labels are inconsistently applied across platforms, as the same IPL photo did not carry any tags on Facebook.

Besides labelling original photos as AI-generated, Meta is also adding labels to those photos that may have been edited using AI features such as Adobe’s Generative Fill. As per reports, even slightly altering a photo using such AI features can be detected by Meta, and flagged as being AI-generated.

How does Meta identify and label AI-generated content?

Currently, Meta has adopted a two-pronged approach to identifying and labelling AI-generated content. Firstly, it requires users themselves to label photorealistic video or realistic-sounding audio that has been generated or modified using AI. This can include an AI-generated video that realistically shows a group of people walking around an outdoor market or a reel with a realistic AI-generated voice-over. If users fail to label such content, Meta warns that they could be penalised.

However, these labelling requirements do not apply to users posting AI-generated content that’s not photorealistic such as a video of an outdoor landscape generated in a cartoon-like caricature.

In addition to having users self-disclose AI-generated content, Meta said that it will also detect AI use on its own. “Any content that contains industry-standard signals that it’s generated by AI will be labeled as “Made with AI.” This includes content that is created or edited using third-party AI tools,” a blog post read.

It could mean that Meta’s detection tools are reading the metadata of photos to determine that they’re AI-generated even if the images were only altered using AI.

The incorrect labelling of original content as AI-generated could have consequences. For instance, it could lead to users dismissing the ‘Made with AI’ label altogether – even if they are being correctly applied. Users have also pointed out that there is no option to uncheck or remove the ‘Made with AI’ label. As a result, many of them are devising workarounds to Meta’s AI labelling initiative such as copying and pasting the image into a blank Photoshop document or uploading a screenshot of the image instead of the original image.