Prof. Soheil Feizi from the University of Maryland’s Reliable AI Lab raised concerns about the effectiveness of Meta’s AI detection system, suggesting it could be easily circumvented by lightweight image processing techniques, leading to high false positive rates.
While Meta’s tool won’t address AI-generated audio and video, users will be asked to label their own content, with potential penalties for non-compliance.
Sir Nick Clegg acknowledged the difficulty in detecting AI-generated text and admitted to the limitations of Meta’s current media policy.
The Oversight Board criticized Meta’s policy on manipulated media as incoherent and lacking justification, particularly in light of a recent ruling regarding a video of President Joe Biden. Despite the criticism, Sir Nick broadly agreed with the ruling and acknowledged the need for updated policies to address the increasing prevalence of synthetic and hybrid content.
Since January, Meta has required political adverts using digitally altered media to be labeled accordingly.
