← sources

How do you trust and/or verify what AI agents did after your prompt?

source: reddit (reddit:AI_Agents)published 2026-09-27 17:25ingested 2026-09-28 02:36

retained text

Of course I know you can read their whole output, but does anyone actually do that?

It seems the productivity you get from using AI is the fact you can ask it to do something, go away, come back and it's done. Huge time saver.

But how do you know that it actually did what you asked, to completion, in a secure manner without doing something else you did not intend?

Take the example of an AI agent installed on a fresh linux server. You ask it to go and set up the box for server use with things such as setting up a firewall, user management etc.

How can you trust that the linux box is now secure? How do you know the AI didn't leave a gaping hole in one area of a secure setup that you would not have missed had you done yourself?

This is one thing that's really holding me back from using AI - I don't trust it to do the actual important tasks where I need to know it is done in a complete and secure manner.

document fetched 2026-09-28 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1wrq154/how_do_you_trust_andor_verify_what_ai_agents_did