Recent AWS outages blamed on AI tools - at least two incidents took down Amazon services

At least two recent Amazon outages were caused by misconfigured AI toolsAmazon asserts incidents were "user error, not AI error"The company has implemented "numerous safeguards"

At least two recent AWS outages were caused by incidents involving Amazon's own AI coding tools, the company has said.

A report by the Financial Times (FT) notes a 13-hour interruption in mid-December 2025 was the result of Amazon's Kiro AI coding agent, which had reportedly decided to delete and recreate the environment.

Although AWS had published an internal report on the cause of the issue, this was never shared publicly, but the FT has obtained the information from four unnamed people familiar with the matter.

Amazon outages caused by internal AI tools

Though Amazon's own AI tools were partly responsible for the outage, the company did stress that "user error, not AI error" was the ultimate cause, attributing the outage to misconfigured access controls.

"The engineers let the AI [agent] resolve an issue without intervention," one of the FT's sources wrote. "The outages were small but entirely foreseeable."

Amazon described this particular incident as an "extremely limited event," but another 15-hour outage in October 2025 had broader implications affecting public apps and websites.

Again, the FT's sources suggest incorrect permissions were to blame, with the AI tools given the same permissions as human workers and its output not given the same approval as would usually be the case with human workers.

Despite very clear dangers, the unnamed sources shared that Amazon is targeting an 80% AI adoption rate among its developers, based on once-per-week usage. A target that could increase as adoption rises.

Speaking about the AI-induced incidents, Amazon wrote: "Following the December incident, AWS implemented numerous safeguards."