Outpaced by AI, why Captchas might disappear
AI Bypasses Captcha Security Measures
“I am not a robot.” In the vast expanse of the Internet, you have almost certainly had to prove your human nature by checking a box, clicking on traffic lights, or deciphering a twisted and scratched-out word. These Captchas, the small guards of the Web, examine you, verify that you are not a machine, and then let you pass. But with artificial intelligence (AI), bots can now deceive this security and sneak onto sites that are forbidden to them.
Captchas no longer seem to be able to stop ChatGPT, according to a report in late July by the specialized site Ars Technica: OpenAI’s tool, through its paid “agent” option, is now quite capable of checking the small box. “This step is necessary to prove that I am not a robot,” the machine then explains calmly.
AI is also winning against image selection Captchas, those that ask you to click on the squares containing specific objects, often from Google Maps. A group of Swiss researchers thus, in September 2024, developed an AI capable of beating the most popular Captcha model “in 100% of cases.” “We took an AI model pre-trained to recognize images and made small adjustments,” explains Andreas Plesner, who worked on the study. “We have shown that anyone could break Captchas using a publicly available artificial intelligence program.”
The rest of this article is reserved for subscribers.
Enjoyed this post by Thibault Helle? Subscribe for more insights and updates straight from the source.
