Sources
Loading...
Additional media
Loading...
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
Critical vulnerabilities in ChatGPT and its plugins have been identified, posing risks of data breaches and account takeovers. Researchers found flaws in encryption, allowing hackers to read private AI assistant chats and potentially hijack accounts on third-party websites.
#FPTech: Hackers can easily read what you say to ChatGPT, other AI services, finds study https://t.co/pjdPHkkXFk
ChatGPT side-channel attack has easy fix: token obfuscation https://t.co/bNJc0W9tYZ
Researchers discover flaw in the encryption of ChatGPT conversations and other AI assistants: the technique allows a network attacker to deduce 29% of the answers, with perfect word accuracy, by observing the transit of data over the internet. OpenAI has already implemented