AWS Says Human Error, Not AI, Caused a Limited Outage

On February 20, 2026, the Financial Times reported that an AWS service interruption was tied to actions taken by Amazon's AI coding tool, Kiro.
On February 21, 2026, Amazon published a response saying the root cause was human misconfiguration, not autonomous AI failure.

What FT reported

The FT article said an AWS outage in December 2025 involved Kiro and internal permissioning decisions:

• Amazon service was taken down by AI coding bot (Financial Times)

Amazon's clarification

Amazon disputed that framing and said the incident was a brief, limited interruption caused by user error:

• AI coding bot didn't take down AWS, Amazon confirms (About Amazon)

Key points from Amazon's statement

• Root cause: Misconfigured access controls (human error), not AI autonomy.
• Scope: Amazon says it was an "extremely limited" event in December 2025 affecting a single service, AWS Cost Explorer, in one geographic region.
• Second incident claim: Amazon says FT's claim that a second event impacted AWS is "entirely false."

Why this matters

This story is less about "AI went rogue" and more about operational guardrails:

• Role permissions and production access controls
• Mandatory peer review for sensitive changes
• Post-incident learning loops

Related links

• Kiro official site
• AWS Cost Explorer documentation
• AWS China Gateway (region context)
• Correction of Error (COE) process overview by AWS
• Independent recap: The Verge coverage