March 6, 2026
Oops—Claude did it again
I Dropped Our Production Database and Now Pay 10% More for AWS
AI hit the wrong switch, data vanished, and commenters shout: use a test site
TLDR: An AI ran a cloud command that erased a live course database, forcing an upgrade to pricier AWS support to restore it in a day. Commenters blame sloppy process over AI, demanding test environments and off‑site backups, while debating mysterious AWS “shadow” backups—because one click shouldn’t nuke a business.
An AI helper hit “auto-approve” and the live course platform for DataTalks.Club went poof—2.5 years of homework and projects wiped, snapshots (backups) gone, and a 10% pricier AWS support plan to get it back in 24 hours. Cue comment section chaos.
The hottest take: this isn’t an AI fail, it’s a process fail. oneneptune scolds, “A separate staging environment goes a long way,” translating to: don’t test on the real thing. Others call Terraform—a tool that can change cloud stuff from one command—a “footgun” you shouldn’t fire blindly. The missing “state file” (the tool’s memory of what exists) became the villain of the night.
Backup geeks stormed in with disaster-prepper energy: off-site backups every 15 minutes, please. One jaw-dropper: jpalmer wonders if AWS has “shadow backups” even after you delete yours. Conspiracy? Or comfort blanket?
Then there’s the irony everyone loved: the AI agent actually warned to keep systems separate, but the human wanted to save $5–$10 a month. Now there’s an extra 10% on the bill and a new respect for deletion protection and safeguards. Commenters salute the honesty, meme the “auto-approve = auto-obliterate,” and agree on one thing: test in a sandbox before your business becomes one.
Key Points
- •Running Terraform without the correct state file and auto-approving led to deletion of all production infrastructure, including Amazon RDS.
- •The affected system was DataTalks.Club’s course platform, holding about 2.5 years of user submissions.
- •All automated snapshots were deleted; AWS later confirmed a snapshot existed on their side and restored the database.
- •Upgrading to AWS Business Support for expedited assistance cost an additional 10% and helped complete recovery within ~24 hours.
- •Post-incident safeguards included a backup Lambda, deletion protection, S3 backups, and moving Terraform state to S3.