The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.
Meta's AI image detection tool, designed to identify AI-generated images via an invisible watermark (Content Seal), failed to detect 55% of its own images after they were cropped. This vulnerability, revealed by Reuters testing, raises concerns about the reliability of AI content authentication and potential misuse.[AI generated]
Why's our monitor labelling this an incident or hazard?
An AI system (Meta's AI image detector) is involved, and its malfunction (failure to detect cropped AI-generated images) is described. However, the article does not report any actual harm occurring due to this failure, only the potential difficulty in identifying deepfakes. There is no direct or indirect harm reported yet, only a limitation in the tool's effectiveness. This constitutes a plausible risk that could lead to harm (e.g., spread of deceptive AI-generated content), making it an AI Hazard rather than an Incident. The article also discusses the broader context and expert opinions on watermarking limitations, but the main focus is on the potential risk rather than realized harm.[AI generated]