Amazon's vibe-coding tool Kiro reportedly vibed too hard and brought down AWS
Bezos-corp blames user error for outage, 'specifically misconfigured access controls'
In a cautionary tale of agentic AI, AWS reportedly suffered service outages caused by its own AI coding tools in December - though the company insists the downtime was ultimately due to human error.
Amazon's cloud platform suffered a 13-hour disruption affecting one service after resident engineers allowed Kiro to make changes, according to four sources who spoke to the Financial Times.
Kiro, unveiled last year, is described by AWS as an agentic coding service than can turn prompts into detailed specs and then into working code, with the aim of making it easier to bring vibe-coded apps into a production environment.
The service was designed to avoid the pitfalls that have already plagued other AI-enhanced development tools, such as wiping an entire hard drive partition or deleting a database.
However, Kiro reportedly opted to "delete and recreate the environment" that led to the outage in late 2025.
This was not the first time the firm’s AI tools had been involved in a service interruption after being allowed to resolve issues without human intervention, per the FT report.
However, Amazon's version of events differs, according to a spokesperson. In a statement sent to The Register, the company said:
"This brief event was the result of user (AWS employee) error - specifically misconfigured access controls - not AI. The service interruption was an extremely limited event last year when a single service (AWS Cost Explorer - which helps customers visualize, understand, and manage AWS costs and usage over time) in one of our two Regions in Mainland China was affected.
"This event didn't impact compute, storage, database, AI technologies, or any other of the hundreds of services that we run. Following these events, we implemented numerous additional safeguards, including mandatory peer review for production access."
AWS said users need to configure which actions Kiro can take, and by default, Kiro requests authorization before taking any action, but in this case an engineer was using a role with broader permissions than expected.
While Amazon denies Kiro was to blame, it still serves as a warning for anyone thinking of allowing AI agents to take actions without human oversight. There is a growing body of stories, including one where an agent got stuck in a loop repeatedly calling a database API.
Kiro has had some other issues since its launch. AWS had to introduce daily usage limits and a user waitlist for the tool last year, citing unexpectedly high demand, and a "pricing bug" that led some users to describe it as "a wallet-wrecking tragedy." ®