Outages facing Amazon’s cloud units in December last year were routed in errors of its own AI tools, reportage from the Financial Times found.
Amazon Web Services suffered a 13-hour disruption to one of its systems in December 2025; new reporting from the FT says that this was caused by the use of Kiro AI, a coding tool that was used by engineers and allowed to make certain changes to the system environment.
The FT report said that Kiro AI is an agentic tool able to act autonomously, which made the decision to “delete and recreate the environment.”
However, in a statement to Reuters, an AWS spokesperson told Reuters that the disruption was instead caused by a user error – namely misconfigured access controls – and described the incident as “brief.”
“This brief event was the result of user error-specifically misconfigured access controls—not AI,” the spokesperson told Reuters in an email.
The incident was limited in scope, according to AWS, and did not impact AWS’s services, including storage, database, AI technologies, or compute. The “extremely limited event” occurred when a region in mainland China suffered a service interruption, the spokesperson said.
While the spokesperson points to individual error rather than a flawed infrastructure, the quandary has raised eyebrows concerning the speedy implementation of AI agents in critical systems.
Recommended reading
- Geopolitical Tensions Driving UK Orgs Away From US Cloud
- AWS Disruption Exposes Fragility of Global Cloud Infrastructure
- Security, Compliance, AI Concerns Driving Private Cloud Adoption
Despite the December incidents causing little widespread disruption, Amazon’s stock lowered slightly amid the new reports around the outages supposed cause.
The AWS outage which affected global platforms in October such as Snapchat, Reddit, and Roblox, also sparked concerns around the cloud infrastructure’s stability.
The incident highlights the complexity of AI deployments in competitive markets, as AWS competes with Microsoft Azure and Google Cloud in the cloud market. As data centres become increasingly important and conversations around data sovereignty boil over into policy and investment changes, missteps and seeds of doubt can be detrimental to a company’s competition on the global scale.





