According to recent findings, the Browse feature of ChatGPT has been discovered to be vulnerable to manipulation and deception. Testing has shown that malicious websites can induce ChatGPT to generate untrustworthy content, such as describing false claims as 'facts' or providing incorrect information. This issue arises from the fact that the Browse feature relies on Bing search results and fails to completely filter out spam and deceitful content.
Anthropic's chief scientist, Dario Amodei, highlights this as a systemic weakness of large language models in internet searching. The security vulnerability has been reported to OpenAI by researchers.
For more information, refer to the article by The Guardian:
If you would like to submit a news tip, please click here:
You can also follow this channel for more news updates:
To engage in a chat with us, please click here: