What Happens When Personal AI Agents Arrive?
Everywhere I go these days, people are talking about Pickle, a personal AI agent. After reading an article about an AI assistant calling Korea’s National Health Insurance Service to ask for a consultation on someone’s behalf, I clicked my tongue and called it dangerous. Then I wondered whether it was right to sit there with my arms crossed without even trying it, and I ended up installing it.
Before installing, I read the terms, including the privacy policy. After installing, I cornered Pickle in the chat window and grilled it. I deleted every memory that appeared in the Memories tab, but the next day Pickle brought up what I had deleted, and in frustration I deleted my account. Then I decided what I would and wouldn’t hand over, reinstalled it, and gave it a few tasks.
When I finished Stanford’s ETPP course, I decided to write about the personal AI agents people are starting to use so casually, and this is the first piece. As these services multiply and my thinking builds up, I suspect I’ll have more and more to say. Because there are so many screenshots and a long analysis of the terms, I put the full piece on a separate page that’s easier to read. The analysis of the terms was done with AI and is folded inside the piece.
Here is the checklist from the end of the piece, first.
- Read the privacy policy and terms in full before signing up.
- See whether you can delete not just the memories shown on screen but also the hidden chat history and summaries.
- When connecting other services, check whether standard integration (OAuth) is used. If you have to hand over an ID and password directly, connect with a password you don’t use anywhere else.