11/07/2026 04:20 - Tecnologia
According to a Reuters analysis published on July 10, 2026, Meta (the parent company of social media giants like Facebook and Instagram) saw its new artificial intelligence image detection tool present certain limitations when identifying images that have been modified. This finding highlights the challenges of content verification in a year of major global elections. Source: El Destape
The study evaluated 40 images created with the generative image model Muse Image, presented this week by the company. While the tool correctly verified 100% of the original images, it failed to identify 55% of those same images once they were cropped to between a third and half of their original size.
Meta explained that its preliminary tool identifies its own images through an invisible digital watermark called Content Seal, which is integrated into each generated image. A watermark is a digital marker used to prove the authenticity and origin of files. However, the company acknowledged that if an image is cropped excessively, this signal can be lost.
The rapid advancement of artificial intelligence brings with it the urgent need to improve tools for identifying deepfakes (highly realistic but fake videos or images created by AI). Companies like Google and OpenAI have also warned that their detection systems are not yet foolproof against image alteration techniques, demonstrating a collective global effort by the tech industry to tackle this issue.
In this regard, Siwei Lyu, a computer science professor at the State University of New York at Buffalo and a researcher in AI image forensic analysis, explained that watermark-based methods can be very effective as long as the signal remains intact. However, modifications such as cropping or intense compression can temporarily reduce their effectiveness.
In March 2026, Meta's Oversight Board (an independent body that reviews content decisions) had already urged the company to redouble its efforts against the proliferation of deceptive content. This type of analysis is fundamental for platforms to continue improving and adapting their technologies, ensuring a safer, more transparent, and reliable digital environment for all users in the future.
Alfredo S. Quiroga
