Instagram is working on labels for AI generated content

Instagram appears to be working on new notices that would identify when AI has played a role in creating content on its platform. 

App researcher Alessandro Paluzzi, who frequently discovers new Instagram features before they’re officially announced or released, has posted a screenshot of a page in the Instagram app that reads “the creator or Meta said that this content was created or edited with AI.” The specific notice notes that in this case it’s an image that’s been “generated by Meta AI,” before giving a brief description of what generative AI is, and how to identify posts that use AI.

The discovery comes shortly after Meta, along with other major AI players including Google, Microsoft, and OpenAI, made commitments to the White House around the responsible development of AI. As well as investing in cybersecurity and discrimination research, one of the commitments included developing a watermarking system to inform users when content is AI generated. 

It’s unclear exactly how automated Instagram’s labeling system will be, and to what extent it will rely on users disclosing when AI has been used to create or edit an image. However, the fact that the notice contains the words “Meta said,” suggests that in at least some cases the tech firm will proactively apply the notice, rather than relying on the honesty of users themselves. A spokesperson for Meta declined to comment to Engadget on the notice, and the company did not immediately respond to The Verge’s request for comment.

Although still in its infancy, we’ve already gotten a taste of what AI-generated misinformation could look like when a picture of the pope in a swagged-out puffy jacket went viral across social media earlier this year. In this case the relatively harmless image was eventually debunked, but it was seen as a warning that simple tools now exist to spread dangerous misinformation if applied to satellite images and political photography.

Meta has recently open-sourced its large language model LLaMA 2, but it’s yet to widely release consumer-facing generative AI features for its products like Instagram. We’ve got a couple of hints of the kinds of features it’s developing, however. In an all-hands meeting in June, CEO Mark Zuckerberg said Meta was developing features like using text prompts to modify photos for Instagram Stories, Axios reported, and app researcher Paluzzi has also spotted signs of an “AI brush” feature for Instagram that could “add or replace specific parts” of images. The Financial Times recently reported that Meta could integrate an AI chatbot ‘personas’ feature into its products as soon as next month.

Beyond Meta, Google has already announced a new tool that should make it easier for users to determine if an image has been AI generated. Its “About this image” feature is launching this summer, and is designed to highlight the first place an image was indexed by the search giant, providing vital clues to its origins.