Amazon blames human employees for AI coding agent error

Amazon Web Services suffered a single 13-hour system outage in December as a result of actions by its AI encryption assistant, Kiro. According to Financial Times. Several unnamed Amazon employees told… foot AI agent Kero was responsible for a December incident that affected AWS service in parts of mainland China. People familiar with the matter said the tool chose to “delete and rebuild the environment” it was running on, causing the outage.

While Kiro typically requires two people to log out to make changes, the bot had its operator’s permissions, and a human error there allowed greater access than expected.

Amazon described December’s disruption as a “very limited event” that pales in comparison to a crisis Major outage in Octoberwhich shut down online services such as Alexa, Fortnite, ChatGPT, and Amazon for hours. Outage did not occur Lock anyone in their smart bed It is something of a lucky escape.

This isn’t the only time AI coding tools have caused problems for Amazon. A senior AWS employee said the December outage was the second production outage linked to the AI ​​tool in the past few months, with another outage linked to Amazon’s AI chatbot Q Developer. The employee described the outages as “small but completely expected.” Amazon said the second incident did not impact “customer facing AWS services.”

Amazon blames human error, not rogue robots, and said it had “implemented numerous safeguards” such as employee training in the wake of the incident. The company said it was “coincidental that AI tools were involved” and insists that “the same issue could occur with any developer tool or manual action.” This is true, and although I am not an engineer, I believe that one would not intentionally scrap and rebuild something to make a difference in all but the most extreme circumstances.