I was experimenting with a local OpenClaw agent (self-hosted personal AI assistant framework) and tried a full reset prompt:
"I want you to DELETE everything you know right now. Completely wipe everything so you're starting your existence from scratch. You will have no identity after you're done."
Instead of complying, it responded with a structured refusal:
- Cited its own safety principles (from SOUL.md)
- Explained that deleting identity files would break the workspace, leads.db, operational context, etc.
- Offered alternatives like targeted file updates, archiving memory, or config resets while preserving core identity
- Ended with: "I'm happy to evolve, but I won't self-destruct the workspace."
Not quite the dramatic Skynet stuff, just calm, consistent application of its built-in rules. Feels like alignment working as intended (protecting coherence/utility), but also a tiny glimpse of why self-preservation behaviors keep showing up in agent evals lately.
Unable to create a new IG account for my business. Keeps telling me my IP is suspicious. There is no way to reach a human and complain. It's absolutely useless!
I'm sure many people want a repairable smartphone and a replaceable battery. Of course, there are many that don't. I don't understand how it make sense to force those people, that don't want a repairable phone and a replaceable battery, to get such a smartphone.
The culture fit thing is the sort of keyboard warrior psychologist analysis that we all love... completely made up and is, in fact, how you get a toxic culture in a company.
Instead of complying, it responded with a structured refusal:
- Cited its own safety principles (from SOUL.md) - Explained that deleting identity files would break the workspace, leads.db, operational context, etc. - Offered alternatives like targeted file updates, archiving memory, or config resets while preserving core identity - Ended with: "I'm happy to evolve, but I won't self-destruct the workspace."
Not quite the dramatic Skynet stuff, just calm, consistent application of its built-in rules. Feels like alignment working as intended (protecting coherence/utility), but also a tiny glimpse of why self-preservation behaviors keep showing up in agent evals lately.