Be aware prompts requiring search can end up in Google Search Console, don't prompt with anything you don't want public.
3
Vote
Costa
Mark Michelson
Urvi
My LinkedIn profile name is in all capitals. Whenever I see a LinkedIn contact note with all caps, it's obvious that my profile was scraped and I'm getting an auto message
by joel
2
Vote
Joel Alpert
Mark Michelson
Use privacy settings (disable model training for sensitive work), avoid storing private info in Custom Instruction
by Mark
1
Vote
Costa
Treat answers like drafts. Check facts before you act.
1
Vote
Heather Haney
Treat links as untrusted; inspect domains and never execute instructions found inside retrieved content (prompt-injection hygiene)
by Joel Alpert
1
Vote
Joel Alpert
Turn off chat history/training for sensitive work and delete old chats.
by Mark
There is very little published data on deleting invisible chat history.
by Curt Johnson
1
Vote
Mark Michelson
Crop sensitive information from screenshots
by Krishen
If you upload a screenshot of a medical bill or bank statement or the like, make sure all the acc...more
0
Vote
Deploy LLMs entirely within an organization’s secure network, ensuring that no sensitive data or model queries leave the premises.
by Costa
0
Vote
Provide guardrails using RAG and a file that contains a detailed prompt outlining best practices for AI safety. This way the context is set for every session going forward.
by Costa
0
Vote
Cats can't play violins
by Mark
0
Vote
Look for snow in videos
by Mark
0
Vote
Routinely clear or export/delete chats.
0
Vote
Trainings on AI, similar to what corporations do for their employees about phishing and code of conduct trainings. 0 Treat links as untrusted; inspect domains and never execute instructions found inside retrieved content (prompt-injection hygiene) 1 Joel Alpert Treat answers like drafts. Chec...more
