News for hackers. Stuff that matters.

A comment on Salesforce study finds LLM agents flunk CRM and confidentiality tests.

Re:Salesforce study finds LLM agents flunk CRM and confident... by anshumankmr on Monday June 16, 2025 @11:02AM (#44290274)
Can this not be solved by RBAC? But I am not sure what all questions were asked and what the setting was, what database was used, what prompts etc.
[ Parent | on HN ]
Re:Salesforce study finds LLM agents flunk CRM and confident... by morgango on Monday June 16, 2025 @11:40AM (#44290644)
Fair question, slightly nuanced answer.

If going against a datasource (like with Retrieval Augmented Generation), yes.

If the information is just part of the context window, no.

[ Parent | on HN ]
Re:Salesforce study finds LLM agents flunk CRM and confident... by anshumankmr on Monday June 16, 2025 @12:10PM (#44290912)
Ideally I would not let anything in the context which is not authorized for the user or the bot is not authorized to do.
[ Parent | on HN ]