Facebook and Instagram Will Soon Label All Fake AI Images

Meta has announced plans to introduce technology that identifies and categorises images generated by other companies’ artificial intelligence (AI) tools. Facebook, Instagram, and Threads will soon be able to label all fake AI images.

While Meta already distinguishes AI-generated images produced by its systems, it aims to extend this capability to images from external sources. It also hopes to catalyze efforts in combating AI-generated misinformation. However, some experts caution that such tools may be circumvented easily.

Facebook and Instagram Will Soon Label All Fake AI Images

Senior executive Sir Nick Clegg outlined Meta’s intention to broaden the labelling of AI-generated content soon. Despite acknowledging the technology’s current immaturity, Clegg emphasized the company’s commitment to fostering industry-wide momentum towards addressing AI-generated fakes.

Nevertheless, Professor Soheil Feizi from the University of Maryland’s Reliable AI Lab warned of the potential limitations of such systems. He suggested that while detectors might identify images generated by specific models, they could be bypassed with simple modifications, leading to false positives and limited applicability across diverse scenarios.

Meta acknowledges that its technology will not extend to audio and video content, which are primary concerns regarding AI-generated misinformation. Instead, users will be tasked with labelling their own audio and video posts, with potential penalties for non-compliance.

Sir Nick Clegg conceded the difficulty in detecting text generated by tools like ChatGPT. He also acknowledged that current policies may not adequately address the proliferation of synthetic and hybrid content.

Recently, Meta’s Oversight Board criticized the company’s policy on manipulated media, deeming it inconsistent and calling for updates. This critique followed a ruling on a video featuring US President Joe Biden, which although not manipulated using AI, sparked debates on Meta’s approach to misinformation.

In response to growing concerns, Meta implemented a policy requiring political advertisements to disclose digitally altered images or videos since January.

Overall, Meta’s efforts recognise the challenges posed by AI-generated misinformation. Yet, the effectiveness of proposed solutions remains uncertain in the face of evolving technology and tactics employed by bad actors.