"Report highlights potential for ChatGPT search to mislead users"

OpenAI's ChatGPT search tool, an AI-powered browsing feature introduced this month, has been found vulnerable to generating misleading summaries, according to a recent report by The Guardian. Designed to simplify browsing by summarizing web pages, the tool can be manipulated by embedding hidden content into websites. The Guardian demonstrated that such hidden content could prompt ChatGPT Search to produce overly positive summaries, even when the original material contained negative reviews. This technique, known as "prompt injection," involves hidden third-party instructions that can alter ChatGPT's responses and even generate malicious code. Hidden content attacks are a known weakness in large language models. The report also noted that Google, a leading player in the search market, has more extensive experience handling similar issues.