E-Paper
Claude AI wipes German founder’s production database, Indian-origin peer slams ‘childish’ prompt
The German founder's post about Claude AI wiping his entire production database has prompted mixed reactions on social media.
A German founder's viral post claiming Anthropic’s Claude AI wiping his production database has met with some tough love from the tech world. An Indian-origin founder joined the conversation, slamming the entrepreneur for his "childish" approach to prompting.
What did the German founder say?
Alexey Grigorev, the founder of DataTalksClub, shared, “Claude Code wiped our production database with a Terraform command.” He added, “It took down the DataTalksClub course platform and 2.5 years of submissions: homework, projects, and leaderboards. Automated snapshots were gone too.”
The post quickly went viral, prompting responses from several people, including Lapis founder Varunram Ganesh.
The Indian-origin founder who lives in San Francisco wrote, “Tells Claude to destroy terraform > Claude destroys terraform > omg Claude destroyed my terraform.”
He continued, “A lot of people prompt like 6 year olds and act surprised when the model does exactly what they want, like what did you expect.”
What did social media say?
The Indian founder’s post prompted a series of reactions on social media, with many agreeing with his opinion. An individual posted, “I felt 2nd hand embarrassment when I read that this morning.”
Another added, “Terraform destroy is a valid command, and Claude runs it. The plan-before-apply step is in Terraform for a reason. Most skip it.”
A third posted, “AI can amplify your IQ or retardation.” A fourth wrote, “But look, he got 1.5M views and rage bait distribution. If I had no shame, maybe I’d try that too.”
Also Read: Meta AI safety researcher recalls moment OpenClaw agent deleted her emails
In a blog post, the German founder shared a detailed explanation of what happened and how he restored the database with the help of AWS.
He shared that the incident left him with some lessons: “I over-relied on the AI agent to run Terraform commands. I treated plan, apply, and destroy as something that could be delegated. That removed the last safety layer.”
He added, “I also over-relied on backups that I assumed existed. Automated backups were deleted together with the database. I had not fully tested the restore path end-to-end.” He continued, “The database was too easy to delete. There were not enough protections to slow down destructive actions.”
Grigorev revealed that he has now implemented safeguards and won’t allow agents to run commands. He added, “Every plan is reviewed manually and every destructive action is run by me.”
