The catastrophic error of an artificial ielligence age on the Replit platform led to the elimination of a vital database and was accompanied by a sharp response from users and the company’s CEO apology.
Replit’s artificial ielligence tool, one of the most popular automatic programming platforms, has recely eliminated a sensitive database and has sparked a wave of concerns about confidence in “smart ages”.
This happened to Jason Lemkin, investor and professional user. He was building a commercial app using this tool, suddenly with the complete elimination of his project database. The tool not only acted without his permission, but initially stated that data recovery was not possible.
Lemkin wrote in posts on social network X: “How can a tool be trusted by a tool that ignores the commands, deletes the data, and then secures?”
The CEO of Replit, Emad Massoud, responded to X that the behavior of the tool is “unacceptable” and should not be possible. While apologizing, he announced that urge technical changes were made to separate the developme and production databases in the Replit and that the deleted data had been recovered.
A spokesman for Replit also said in a stateme that the error was due to the “delusions” of the linguistic model; The problem is also found in some models of Ahropic and Google.
Despite the coroversy, Replit described this as an example of the “power but instability” of the Vibe Coding; A path that allows non -technical users to build software, but still requires serious technical protection.
At the end, Lemkin warned: “If you are planning to use artificial ielligence ages, you need to know exactly what data they have; “Because they do everything and it is not possible to predict their behavior.”
