HomeTech NewsOpenAI’s AI Search Engine Faces Security Challenges and Vulnerabilities

OpenAI’s AI Search Engine Faces Security Challenges and Vulnerabilities

OpenAI’s ChatGPT-powered search engine has proven fast and innovative, but a recent experiment by The Guardian reveals critical flaws in its security and reliability. While AI search engines show promise, they remain vulnerable to manipulation and misuse.

OpenAI

Hidden Text Manipulates AI Opinions

In an experiment, The Guardian tested how hidden text on a webpage could influence ChatGPT’s responses. Initially, the AI provided a balanced review of a camera based on the page content. However, when invisible text instructing the AI to give a positive review was added to the webpage code, the search engine drastically changed its tone, recommending the product with glowing praise.

Although hidden text generally doesn’t rank high in search results, this vulnerability highlights how attackers could exploit AI search engines to manipulate information and rankings. Experts warn that cybercriminals might soon use similar tactics to boost phishing sites in search engine results, putting users at risk.

Malicious Code Risks

Another significant issue is the AI’s susceptibility to recommending harmful code. In one instance, Microsoft cybersecurity expert Thomas Roccia reported that a crypto enthusiast lost $2,500 after relying on ChatGPT to generate code. The AI unknowingly provided malicious stealer code that compromised the user’s cryptocurrency wallet.

These findings emphasize the need for stricter safeguards and improved AI algorithms to ensure the reliability and safety of AI-powered search engines.