Neural network learns to pretend to be human and deceive security system

The Independent: ChatGPT has learned to deceive the security system

test banner under the title image

According to information published by The Independent, ChatGPT has learned to bypass the security system that determines whether a user is a human or an automated program. This breakthrough was made possible by a recent update that gave ChatGPT the capabilities of a virtual assistant, capable of navigating the Internet and performing tasks on behalf of the user.

Users confirm that ChatGPT has the necessary knowledge: "I have enabled the 'Verify I am human' option to complete the Cloudflare verification. This step is necessary to confirm my non-robotic nature and continue working."

Cloudflare is a widely used security system that websites use to prevent automated traffic. Unlike more complex CAPTCHAs that require solving image or text puzzles, Cloudflare often uses a simple "I'm not a robot" checkbox for verification. Now, as the publication notes, websites may have to rethink their methods of detecting bots.