Meta unveils plans to label AI-Generated content on its platforms

Meta, the parent company of Facebook, Instagram, and Threads, has recently unveiled plans to implement comprehensive labeling for all AI-generated content across its platforms. This decision follows the discovery by Reverse Engineer Alessandro Paluzzi in August of the previous year, revealing Instagram’s intentions to introduce labels identifying posts created or modified using generative AI tools. The announcement was made through an in-app message shared by Paluzzi, shedding light on the forthcoming change.

In an official statement, Nick Clegg, the President of Global Affairs at Meta, articulated the company’s commitment to collaborating with industry partners to establish common technical standards for identifying AI-generated content, encompassing video and audio formats. The primary objective behind this initiative is to delineate a clear distinction between content produced by humans and that generated through artificial intelligence, thereby fostering transparency on Meta’s platforms.

Clegg emphasised that Meta currently employs measures to address photorealistic images generated using its AI feature. These measures include incorporating visible markers that users can observe on the images, along with the utilization of invisible watermarks and embedded metadata within image files. The intention is to provide users with tangible indicators of AI involvement in content creation.

The delineation between AI and human-generated content is pivotal, according to Clegg, as the lines between the two become increasingly blurred. Users encountering AI-generated content for the first time have expressed a desire for transparency regarding this new technology.

To meet this demand, Meta plans to introduce “Imagined with AI” labels for photorealistic images created using its Meta AI feature. The company is also collaborating with industry partners to establish technical standards for labeling content generated with tools from other companies.

Furthermore, Clegg revealed that Meta is developing industry-leading tools capable of identifying invisible markers at scale, specifically aligning with the “AI generated” information outlined in the C2PA and IPTC technical standards. This development aims to enable Meta to label images generated by tools from prominent companies such as Google, OpenAI, Microsoft, Adobe, Midjourney, and Shutterstock.

Looking ahead, Meta intends to implement these labeling capabilities in all languages supported by each app in the coming months. The phased approach will coincide with a period of global significance, as several important elections are scheduled.

Meta anticipates gaining insights into user interactions with AI content, preferences for transparency, and the evolution of AI technologies. The findings during this period will inform industry best practices and Meta’s ongoing approach to AI-generated content labeling.