New research has revealed that ChatGPT search can be spoofed to mislead users

ChatGPT Search, an AI-powered search engine The live broadcast has started This month, they could be tricked into generating completely misleading summaries, a British newspaper reports The Guardian newspaper has found.

ChatGPT’s search feature aims to make browsing faster by doing things like summarizing a web page’s product reviews. But The Guardian found that it could make ChatGPT ignore negative reviews and produce “entirely positive” summaries by inserting hidden text in Websites you created. ChatGPT research can also be done to release malicious code using this method.

Such hidden text attacks are Known pitfalls of large language modelsbut this appears to be the first time it has been offered on a live AI-powered search product. The Guardian newspaper indicated that Google, the leader in the field of search, has more experience in dealing with similar problems.

OpenAI did not comment on this specific incident when contacted by TechCrunch, but said it uses a variety of methods to block malicious websites and is constantly improving.

TechCrunch has an AI-focused newsletter! Register here Get it in your inbox every Wednesday.