The Kiro autonomous code-writing tool could have caused a prolonged AWS outage: what is known about the incident?
In December, a 13-hour outage occurred in the Amazon Web Services (AWS) infrastructure, which, according to the Financial Times, could have been linked to an internal tool with artificial intelligence, Kiro. It is about a code-writing system capable of performing autonomous actions on behalf of the user.
What is known
According to four sources familiar with the situation, engineers used Kiro to make changes to the environment, after which the system allegedly initiated a "deletion and re-creation of the environment", leading to a prolonged outage that mainly affected China.
Amazon denies that the incident was caused by AI autonomy. The company stated that it was a "coincidence", and the root cause was a user error — incorrectly configured access permissions. According to the official position, the Kiro artificial intelligence tool by default requests confirmation before performing any actions. However, in this case, the employee had broader powers than expected, which allowed the critical operation to be performed without additional control.
The company emphasizes that the incident was limited and related only to the AWS Cost Explorer service in one of the 39 geographic regions. According to Amazon, the outage did not affect computing resources, data storage, databases, or artificial intelligence services. The company also stated that no client inquiries regarding this incident were received.
After the incident, AWS introduced additional security measures. Meanwhile, FT sources claim that this is at least the second case in recent months where the company's artificial intelligence tools are at the center of service disruptions.
Earlier, in October, AWS experienced a larger-scale outage that affected the operation of Alexa, Snapchat, Fortnite, and Venmo. At that time, the cause was a software error in automation.
Source: Financial Times
