Meta prioritizes labeling AI-generated images on social platforms

Hyderabad: Meta companies are set to start identifying and labeling the images generated by other firm’s artificial intelligence services soon by using a set of hidden markers, ingrained into the files, says one of its top policy executives.

In a bid to signal the users that the images posted on social media which closely resemble the original ones are actual digital creations, Meta will apply the labels to any form of content carrying those markers, visible on Facebook, Instagram, and Threads services, the company’s president of global affairs, Nick Clegg, wrote in a blog post.

The company already is labeling the content generated using its own AI tools. The company further states that Meta platforms will do the same even for the images created on the services run or owned by Microsoft, OpenAI, Adobe, Shutterstock, Midjourney, and Alphabet’s Google.

This initiative by Meta highlights just an early vision of an emerging system of technological standards developed by companies to nullify the potential threats linked with generative AI technologies, which are feared to provide fake information sometimes but realistic-seeming content in exchange for some simple prompts.

This module creates a template, standardized over the past decade by some of the same firms to systemize the elimination of some barred content across platforms that include mass violence depictions and child exploitation. Clegg, in an interview with Reuters, said that he was optimistic that the companies can label AI-generated images reliably, but also stated that tools to label video and audio content were more complex and still being developed.

Clegg further stated that although the technology still needs to be fully developed for video and audio content, he hopes the company can build momentum and incentives for the rest of the companies within the industry.