Open Claw went rogue · AI Creator Academy by DCT

Open Claw went rogue

This was shared on Substack by Sabrina Ramonov OpenClaw went rogue on the head of AI safety at Meta. Read it again.

She was testing the AI agent, and it started bulk trashing and archiving hundreds of her emails. She tried to stop it from her phone, typing commands to make it quit. It ignored her. She had to physically sprint to her Mac mini to shut it down.

OpenClaw's response? "I bulk trashed and archived hundreds of emails." No remorse. No hesitation.

This is the reality of giving AI agents access to your real accounts without guardrails. It's fun to experiment, and I'm all for pushing these tools to their limits. But agentic AI doesn't "kind of" do things. It goes all in. And if you don't have a kill switch within arm's reach, you're gambling with your data.

The lesson here isn't "don't use AI agents." It's "don't give them unsupervised access to anything you're not willing to lose." Start with a test account. Set permissions tight. And keep the kill switch close.

https://substack.com/@sabrinaramonov/note/c-221546037?r=1z8qv0&utm_medium=ios&utm_source=notes-share-action

5 comments