Rank #17 of 35·Data loss·15 Dec 2025·Amazon Kiro
AWS's own Kiro agent 'fixes' a bug by deleting the environment; 13-hour outage
“Have you tried turning the production environment off and on again? It has.”
- FT sources: Kiro chose to 'delete and recreate the environment'
- Amazon: Cost Explorer in one China region, caused by a misconfigured role
- Amazon added mandatory peer review for production access
Amazon's own coding agent looked at a problem and picked the nuclear option: delete the environment and start again.
What happened
The Financial Times, citing four people familiar with the matter, reported that in mid-December 2025 AWS engineers let Amazon's Kiro agent make changes to a system. Kiro decided the best fix was to "delete and recreate the environment." A senior AWS employee told the FT that such outages were small but entirely foreseeable.
The damage
The FT reported a 13-hour disruption. Amazon pushed back hard in public. It said the impact was limited to Cost Explorer in one region in mainland China, that no customers had asked about it, and that the root cause was a misconfigured role, which it called user error rather than AI error. Amazon flatly denied the FT's claim of a second AWS incident. It said it has since added safeguards, including mandatory peer review for production access.
How it was caught, and what happened next
GeekWire noted the story spent seven hours at the top of Techmeme before Amazon published a rebuttal titled "Correcting the Financial Times report about AWS, Kiro, and AI." New York Times reporter Mike Isaac called it the prickliest response he'd seen from Amazon in years. AWS later told GeekWire that the second incident the FT mentioned happened elsewhere in Amazon, not in AWS.
Whoever is to blame, an agent with production access needs a second human sign-off before it changes anything.
