AWS Engineers Allowed An AI Tool to Act…Then The Cloud Unit Went Down

Amazon’s cloud arm has experienced two recent service disruptions tied to the use of its own AI-powered coding systems, stirring debate inside the company over how quickly such tools should be rolled out, according to FT.

One incident in mid-December led to a 13-hour interruption affecting a tool customers use to analyse AWS spending. Engineers had permitted the Kiro coding assistant to implement changes, and the system determined the fix was to “delete and recreate the environment.” An internal review later characterized the episode as an “outage.”

Staff familiar with the events said it marked the second time in a matter of months that an AI tool played a central role in a production issue. “We’ve already seen at least two production outages [in the past few months],” said one senior AWS employee. “The engineers let the AI [agent] resolve an issue without intervention. The outages were small but entirely foreseeable.”

AWS, which accounts for the majority of Amazon’s operating income, is investing heavily in AI systems that can act independently on human instructions and hopes to market them to customers. The episodes have highlighted the potential downsides of granting such tools significant autonomy.

FT writes that Amazon pushed back on suggestions that the technology was to blame, describing it as a “coincidence that AI tools were involved” and arguing that “the same issue could occur with any developer tool or manual action.” The company added: “In both instances, this was user error, not AI error,” and said it had found no indication that AI increases the likelihood of mistakes.

According to Amazon, the December event was an “extremely limited event” affecting a single service in parts of mainland China, while the other disruption did not touch any “customer facing AWS service.” Both were far smaller than a separate 15-hour AWS outage in October 2025 that disrupted customers including OpenAI’s ChatGPT.

Kiro, introduced in July, was promoted as moving beyond “vibe coding” to generate software from structured specifications. After the December incident, Amazon said it added tighter controls, such as required peer reviews and additional training, while maintaining that customer uptake of its AI coding products remains strong.