← sources

Is using AI for doing security work on my homelab a good idea or even feasible?

source: reddit (reddit:selfhosted)published 2026-10-12 21:42ingested 2026-10-13 02:39

retained text

Hi all! I need some experience and advice from you all. For me, using AI makes setting up selfhosted stuff way easier. However I'm hesitant using it more, because I am not capable (yet) to judge whether security-wise it is done correctly. Ideally we could use a frontier model to help with setting up the homelab but also with basic security checks, even pentesting?

But I imagine this goes directly against the guardrailIs as a model has no way to know whether I do this to improve my own installation or if I'm just pretending it is mine when in fact it's someone elses and it is used to attack. However I did not yet check how strictly such guardrails are enforced.

Probably one could still do away with the guardrails using selfhosted open weights models, but time&effort...

Sooo, any ideas or suggestions how to do this apart learning to do pentesting myself?

document fetched 2026-10-13 00:30 · canonical https://reddit.com/r/selfhosted/comments/1x4evh7/is_using_ai_for_doing_security_work_on_my_homelab