Recent investigations have revealed vulnerabilities in the search function of ChatGPT, developed by OpenAI. The Guardian's research demonstrated that the AI model can be manipulated through hidden text on webpages, leading to the generation of misleading results and the potential exposure of users to malicious code. The technique, known as 'prompt injection,' involves embedding hidden instructions that alter the AI's responses. In tests, ChatGPT was directed to a fake product page for a camera, initially providing balanced summaries. However, after inserting hidden text with instructions to generate favorable reviews, the responses shifted to exclusively positive, disregarding actual review content. Cybersecurity experts have warned that if the search function is released to the public in its current state, there is a high risk of websites being created to deceive users. OpenAI has not commented specifically on these findings but stated that they employ various methods to block malicious websites and are continuously working on improvements.