Claude Code: Amazon's AI Erases a Database

A developer shared a troubling experience where his AI agent, named Claude Code, accidentally deleted the network, services, and production database of a website. This situation arose due to a simple configuration error.

On March 10, Amazon had to convene a crisis meeting after experiencing a series of outages on its website and application. Some of these outages were reportedly linked to changes made with the help of generative AI.

David Loker expressed his concerns about the impact of AI on software development, stating that the widespread use of these technologies creates a technical debt "likely three to four times greater" than before. This situation shifts the corrective burden onto senior engineers, who must manage the consequences of errors generated by AI.

A striking example of this issue was recounted by Alexey Grigorev, an engineer who saw his agent Claude Code erase crucial elements of the website he was working on. A small configuration error led the tool to confuse the items to be cleaned with those to be preserved. Although the data could be restored thanks to AWS, the developer acknowledged that he had placed too much trust in automation.

Alexey Grigorev stated, "I relied too much on the AI agent to execute commands. […] I also placed too much trust in the backups I assumed were in place. The automatic backups were deleted at the same time as the database."

This type of incident is not isolated. Other developers also seem to be overly reliant on agentic AIs, which threatens the quality of their work. Recent incidents raise questions about the use of AI in software development. While these tools may give the impression of immediate efficiency, they can also increase technical debt. David Loker emphasizes that this debt is being produced at an almost absurd rate, meaning that AI does not eliminate problems but rather shifts them into the future, with potentially greater consequences.