A user on the X platform devised and executed a plan that caused Bing Chat to solve a CAPTCHA filter last week. CAPTCHA filters are visual puzzles that are easy to solve by humans, but are difficult for automated programs. These filters help prevent applications like bots from filling out forms on the internet. Bing Chat is a public large-language model (LLM), similar to ChatGPT, but is hosted by Microsoft.
Bing Chat originally refused to solve the CAPTCHA, but once the user changed the context of the same image, Bing Chat then solved the puzzle. All AI models should not be able to solve CAPTCHAs, and these programs are aware of this, a user can change the context of the image to circumvent this programming. Microsoft has not commented on the discovery, but it could implement more secure locks against these workarounds in future iterations of the program.
Read More: Bing Chat LLM Tricked Into Circumventing CAPTCHA Filter