Report: Amazon’s AI bots have been behind multiple AWS outages
Amazon’s AI tool Kiro, which launched in July and can code autonomously, was behind a 13-hour interruption to Amazon Web Services in December, according to reporting by the Financial Times.
The FT reports that the company’s AI tools have caused AWS service disruptions at least twice in recent months.
In the December outage, which Amazon called an “extremely limited event” that did not have an impact on customer-facing service, engineers allowed Kiro to make changes and the tool opted to “delete and recreate the environment.”
Amazon has a closely tracked internal target that 80% of its developers use AI to code once a week, employees told the FT. The company says the December incident was a “user access control issue” and not an issue with Kiro’s permissions.
AWS accounted for 57% of Amazon’s operating profit in 2025. In December, following a larger outage months earlier, AWS and Google announced a partnership to attempt to prevent massive network outages.
Update, February 20, 5:50 p.m. ET: In a statement to Sherwood News, an AWS spokesperson disputed the report, writing:
“These brief events were the result of user error — specifically misconfigured access controls — not AI. The December service interruption was an extremely limited event 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. We are not aware of any related customer inquiries resulting from this isolated interruption. Following these events, we implemented numerous additional safeguards, including mandatory peer review for production access, enhanced training on AI-assisted troubleshooting, and resource protection measures. Kiro puts developers in control — users need to configure which actions Kiro can take, and by default, Kiro requests authorization before taking any action.”
In the December outage, which Amazon called an “extremely limited event” that did not have an impact on customer-facing service, engineers allowed Kiro to make changes and the tool opted to “delete and recreate the environment.”
Amazon has a closely tracked internal target that 80% of its developers use AI to code once a week, employees told the FT. The company says the December incident was a “user access control issue” and not an issue with Kiro’s permissions.
AWS accounted for 57% of Amazon’s operating profit in 2025. In December, following a larger outage months earlier, AWS and Google announced a partnership to attempt to prevent massive network outages.
Update, February 20, 5:50 p.m. ET: In a statement to Sherwood News, an AWS spokesperson disputed the report, writing:
“These brief events were the result of user error — specifically misconfigured access controls — not AI. The December service interruption was an extremely limited event 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. We are not aware of any related customer inquiries resulting from this isolated interruption. Following these events, we implemented numerous additional safeguards, including mandatory peer review for production access, enhanced training on AI-assisted troubleshooting, and resource protection measures. Kiro puts developers in control — users need to configure which actions Kiro can take, and by default, Kiro requests authorization before taking any action.”
