Facebook and Instagram to Label All Fake AI Images

Meta, the parent company of Facebook and Instagram, has announced plans to introduce technology capable of detecting and labelling images generated by external artificial intelligence (AI) tools on its platforms.

Currently, Meta already labels AI-generated images produced by its own systems, but this new initiative aims to extend this labelling to images generated by other companies’ AI technologies. The technology will be integrated into Facebook, Instagram, and Threads.

Sir Nick Clegg, a senior executive at Meta, highlighted in a blog post that the company intends to expand the labelling of AI-generated images “in the coming months.” Despite acknowledging that the technology is not yet fully mature, Clegg expressed Meta’s desire to lead the way in addressing AI fakery and to incentivize the industry to follow suit.

However, Professor Soheil Feizi from the University of Maryland’s Reliable AI Lab cautioned that such detection systems could be circumvented easily. He explained that while detectors might flag specific images generated by certain models, they could be evaded through simple manipulations and might produce a high rate of false positives, limiting their effectiveness across various applications.

Meta also admitted that its AI detection tool would not be applicable to audio and video content, despite concerns primarily revolving around these media types. Instead, Meta plans to rely on users to label their own audio and video posts, with potential penalties for non-compliance.

Regarding the challenge of identifying text generated by AI, such as ChatGPT, Sir Nick Clegg conceded that it would be impractical to test for this.

Meta’s announcement comes amidst criticism from its Oversight Board regarding its policy on manipulated media. The Board deemed Meta’s current policy “incoherent” and suggested updates to better address synthetic and hybrid content.

Despite these challenges, Meta has already implemented policies requiring political advertisements to indicate the use of digitally altered images or video since January.

Meta’s efforts reflect a growing recognition of the need to address the proliferation of AI-generated content and the challenges it poses in maintaining the integrity of online platforms.