AI or Access Error? AWS Outage Sparks Debate

A section of Amazon Web Services (AWS) experienced a disruption lasting nearly 13 hours last December. A report by the Financial Times claims the prolonged outage was triggered by the actions of the company’s own artificial intelligence (AI) tool. The news was also reported by Engadget.

According to the report, engineers had assigned a minor technical issue to an ‘agentic’ AI tool named Kiro. However, the tool—capable of making autonomous decisions—allegedly chose to delete and rebuild the entire system environment automatically, resulting in the extended service disruption. The outage affected customer services in a region of mainland China.

Amazon, however, has rejected claims that the AI tool was to blame. The company stated that the incident was caused by “user error” rather than any fault of the AI system. In a statement, an AWS spokesperson said, “This brief incident was the result of user error—specifically misconfigured access controls—not AI.”

The company further explained that Kiro, by default, seeks permission before executing tasks. In this case, the employee involved reportedly had broader permissions than expected, pointing to an issue with access control settings. Amazon maintains that the AI tool’s involvement was “purely coincidental,” arguing that a similar issue could have occurred with any developer tool or even through manual intervention.

The Financial Times report noted that this was at least the second service disruption in recent months allegedly linked to Amazon’s AI tools. Another AI tool, called Q Developer, was reportedly involved in a separate system disturbance. A senior AWS employee described the incidents as “minor but entirely predictable.”

It is worth noting that in October 2025, another AWS outage lasting 15 hours disrupted services such as Alexa, Snapchat, and Fortnite. At the time, the company attributed the disruption to a flaw in its automation software.

DBTech/BMT/OR