Amazon’s cloud division suffered outages last year that were reportedly triggered by its own AI systems.
One 13-hour disruption in December occurred when an AI agent autonomously deleted and rebuilt part of its environment.
AWS underpins large parts of the internet and has faced wider reliability concerns.
An October incident temporarily knocked dozens of websites offline and highlighted dependence on a few major providers.
Amazon said only one smaller event affected customer services.
The company blamed “user error” and misconfigured access controls rather than the AI itself.
It added that new safeguards now require peer review for production access.
The claims come as Andy Jassy oversees major job cuts.
Amazon announced 16,000 layoffs in January after 14,000 the previous autumn.
Jassy has said AI will improve efficiency but not directly replace staff.
Some cybersecurity specialists dispute the company’s explanation.
They argue AI agents can act quickly without fully understanding wider consequences.
Unlike humans, they may lack context about system importance or downtime costs.
Amazon insists the interruptions were limited and did not affect core services such as compute or storage.
The debate adds to broader questions about how safely companies can deploy autonomous AI inside critical infrastructure.
