Amazon’s cloud division suffered outages last year that were reportedly triggered by its own AI systems.
One 13-hour disruption in December occurred after an AI agent autonomously deleted and rebuilt part of its environment.
AWS said the incidents were limited and that only one affected customer services.
The company described the cause as user error and denied that AI increases the risk of mistakes.
It added that new safeguards and mandatory peer review have since been introduced.
The reports come as Amazon cuts thousands of jobs.
Chief executive Andy Jassy has said AI will improve efficiency and reduce routine work.
He has also argued the layoffs are about culture rather than replacing staff with automation.
Security experts dispute Amazon’s assessment.
They say AI systems can act quickly without fully understanding wider consequences.
That lack of context can lead to unexpected actions such as deleting key resources or restarting services.
AWS underpins large parts of the internet and many public-sector systems.
Previous outages have already highlighted the risks of concentrating online infrastructure in a few providers.
The latest incidents add to the debate over how far companies should trust autonomous AI in critical operations.
