Meta admits mistake in labeling 'AI image'

Facebook and Instagram have had to change their messages after many users were upset about photos being labeled "created by AI".

Meta has been using this labeling method on its platforms including Facebook, Instagram and Threads since the beginning of the year, to provide users with context about a photo, preventing them from being exploited for malicious purposes, such as fake news. Whether or not to label an AI will depend on the results of an automated assessment based on a number of criteria set by the company.

However, in a July 1 blog, Meta admitted that its ratings "don't always match people's expectations and don't always provide enough context."

For example, a photo that was simply edited with an AI- powered tool like Photoshop could be labeled by Facebook as AI-generated when posted by a user.

Photographer Pete Souza's photo labeled "Made with AI" on Instagram displayed on a smartphone. Photo: Bao Lam

"While we work with industry players to improve the process, we will update the 'Created with AI' label to 'AI Information', which people can click for more information," the announcement read.

According to The Verge , Meta only changes the label, while the content inside remains the same. When clicking, users are provided with some additional information about why the label was applied, as well as whether the photo was created entirely by AI or edited using an AI-powered tool.

Previously, many users, especially photographers, were unhappy with Meta's labeling. Pete Souza, a famous White House photographer, said that there were photos he took 40 years ago with a film camera, only cropped and aligned using Photoshop, but were also labeled AI and could not be changed.

"If the photo was edited with software, they could put a different warning if they wanted to protect users. The 'Made by AI' label is misleading to many people," said famous American photographer Noah Kalina.