Is your approach inherently secure because you are not getting it to process downloaded files, only text from Obsidian? I have been reading articles about prompt injection, malicious commands in PDF files and on webspages. Is there a risk that with file access, the AI can do some serious damage to your local machine?
Is your approach inherently secure because you are not getting it to process downloaded files, only text from Obsidian? I have been reading articles about prompt injection, malicious commands in PDF files and on webspages. Is there a risk that with file access, the AI can do some serious damage to your local machine?