Claude Code wipes out a production database
xcancel.com"People letting an LLM agent rip on a production system get what they deserved" - I fixed the title.
Funnily enough, they didn't let Claude rip:
They actually reviewed the agent's suggestion before running it. It's just that they accepted it anyway because they didn't understand Terraform enough to realise what it'd do. The agent didn't either, and didn't have enough security mindset to do a `terraform plan -destroy` beforehand, which would have revealed the mistake.The agent kept deleting files, and at some point, it output: “I cannot do it. I will do a terraform destroy. Since the resources were created through Terraform, destroying them through Terraform would be cleaner and simpler than through AWS CLI.” That looked logical: if Terraform created the resources, Terraform should remove them. So I didn’t stop the agent from running terraform destroy.>So I didn’t stop the agent from running terraform destroy.
So they did let it rip.
Spin from AI people...we are getting close to AGI...its every day more human...
Didn't someone recently experience a similar fubar?
Similar, not the same: https://news.ycombinator.com/item?id=46103532