Amazon’s cloud division reportedly suffered outages last year after internal AI software made critical changes to its own systems.
One disruption in December lasted about 13 hours after an AI agent deleted and rebuilt part of its environment.
The incidents have renewed scrutiny of Amazon’s rapid AI rollout.
They also come as the company cuts thousands of jobs while promoting automation.
Chief executive Andy Jassy has said AI will improve efficiency but not simply replace staff.
Amazon disputes the claim that AI caused the failures.
It says the problem was human misconfiguration and only one event affected customer services.
The company added that new safeguards and peer-review rules are now in place.
Cybersecurity specialists question that explanation.
They argue AI agents can act without fully understanding the wider impact of system changes.
Unlike human engineers, automated tools may not pause to reassess risky commands.
The outages were limited compared with larger AWS failures that have previously taken major websites offline.
However, they highlight the growing risks as critical internet infrastructure becomes more dependent on complex AI-driven operations.
