Those “I am not a robot” website tests? Robots can now pass them. ChatGPT can now successfully bypassing the security step offered on many websites to prevent malicious activities like data scrapping often done by bots. OpenAI, the company behind the generative AI chatbot, launched a new model called ChatGPT Agent earlier this month—the “agent” can navigate websites to handle online tasks like ordering supplies or booking appointments. Users have since shared their experiences with it, with one Redditor demonstrating how ChatGPT can bypass Cloudflare’s security check. “Now I’ll click the ‘Verify you are human’ checkbox to complete the verification,” the ChatGPT Agent explained in this scenario, adding that, “this step is necessary to prove I’m not a bot.” (Yes, that’s a bot telling you it’s not a bot.) OpenAI has maintained, however, that human users will always have full control, stating that the bot will always “request permission before taking actions of consequence.” “You can easily interrupt, take over the browser, or stop tasks at any point,” OpenAI said in a press release. The agent also demonstrated it can also bypass more complicated CAPTCHAs featuring images and puzzles by deceiving a human into thinking it was blind, according to a developer livestream.
Read it at The Independent
