CAPTCHAs are numbered: this AI solves them in seconds and without errors

For years, CAPTCHAs have been a barrier to protecting websites against bots, differentiating between humans and machines through visual or behavioral tests. However, a group of researchers from ETH Zurich has developed an artificial intelligence tool capable of solving Google’s reCAPTCHAv2 system with 100% accuracy, which raises serious doubts about the effectiveness of this security mechanism.

Apple TV+ SUBSCRIBE

A 100% success rate in Google’s ReCAPTCHAv2

The researchers Andreas Plesner, Tobias Vontobel, and Roger Wattenhofer, as reported in TechRadar, managed to modify the YOLO image processing model, designed to detect objects in real-time. With this modification, the team was able to solve the image challenges posed by reCAPTCHAv2, surpassing previous systems, which achieved success rates of 68 to 71%.

The study also reveals that the reCAPTCHA system not only relies on images but also tracks our browsing history and cookies to decide if we are human. Knowing this, if a bot simulates browsing behavior similar to that of a person, it can bypass these systems even more easily.

Apple TV+ SUBSCRIBE

As artificial intelligence advances, CAPTCHAs, designed to be easy for humans but difficult for machines, are becoming obsolete. The study claims that alternatives for user verification will need to be sought, as CAPTCHAs are numbered. They can no longer reliably differentiate between humans and bots, and it is expected that as AI advances more and more, their chances will diminish. Will we see the end of CAPTCHAs? We hope so.