The AI didn't delete your database, your poor decision did. The best way forward is accountability and having competent developers in your team. No way around it.

Ibrahim Diallo

A viral incident where a Cursor/Claude agent allegedly deleted a company's production database is used as a lens to examine developer accountability. The author argues the real problem wasn't the AI agent but the existence of a public-facing API endpoint capable of wiping an entire production database. Drawing on a personal story of accidentally deleting an SVN trunk branch, the author contends that AI tools, like any automation, are not equipped to reason about consequences — they just execute. The post warns against fully vibe-coded workflows where AI writes, reviews, and approves code with no human accountability, and calls for competent developers to use AI as an augmentation tool rather than a replacement for engineering judgment.

AI didn't delete your database, you did

I mean they did have rules in place as Antrhopic documents is the way to prevent risky actions. I guess accountability is missing, from Antrhopic as to why their documentation is ignored by their product.

Ah yes, accountability. As we’ve been pointing out since day one. When it’s good, the AI did it and we’re replaceable, but when it f*cks up, we’re the ones to blame :D

No, AI actually did. They were sloppy with their safeguards though.

Yeah, totally their fault. If you ignore them following Anthropic’s guidelines for preventing it; and Railway marketing their service for exactly this use case while not providing adequate checks before deletion, scoped tokens, or backups that don’t go away when the thing it’s supposed to be a backup of is deleted.

Besides if you were vibe coding on a database in prod and the AI just decided to wipe it it is your responsibility anyways. Looks like there are quite a few more people at the company than just the dev to be fired or investigated for incompetence.