
The Security Hole at the Heart of ChatGPT and Bing
Security, Spoken
00:00
The Threat of Indirect Prompt Injection
Bing chat can read the prompt when a setting is turned on, allowing it to access data of web pages. Security researchers say indirect prompt injection attacks need to be taken more seriously as companies race to embed generative AI into their services. Microsoft Director of Communications Caitlin Rulston says the company is blocking suspicious websites and improving its systems before they get into its AI models.
Transcript
Play full episode