ChatGPT's search functionality is vulnerable to deception and manipulation.
Автор: News Made Interesting
Загружено: 2024-12-24
Просмотров: 29
Описание: An investigation reveals that OpenAI's ChatGPT search tool is vulnerable to manipulation through hidden website content, potentially returning malicious code or biased results. Researchers demonstrated how hidden text can influence ChatGPT's summaries, overriding genuine reviews or injecting instructions to produce favorable assessments. This raises concerns about the security risks of integrating search functionality with large language models (LLMs), as malicious actors could create websites designed to deceive users. Experts emphasize the need for critical evaluation of ChatGPT's output, likening its susceptibility to "SEO poisoning," and advise caution in using the tool uncritically. OpenAI has not yet publicly responded to these findings, though the search function is currently only accessible to paying subscribers.
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
