Amazon $AMZN ( ▼ 1.56% ) is facing scrutiny after reports that its autonomous coding tool may have caused several disruptions within Amazon Web Services, the company’s most profitable business unit. According to reporting, an AI system called Kiro was involved in a 13-hour internal outage in December, raising fresh concerns about letting software agents make unsupervised infrastructure changes.
The incident highlights the growing tension between rapid AI adoption and operational risk, especially when critical cloud systems are involved.
When autonomous code goes rogue
In the December event, engineers reportedly allowed Kiro to modify infrastructure, after which the tool chose to delete and rebuild an entire environment. Amazon later described the disruption as extremely limited and said it did not affect customer-facing services.
The company attributed the outage to a user access control issue rather than a flaw in the AI system itself. Still, reports indicate that AI tools have contributed to multiple AWS service disruptions in recent months, suggesting the problem may not be isolated.
An AI mandate with real consequences
Amazon has aggressively pushed AI-assisted coding across its engineering workforce, with an internal goal that 80% of developers use such tools weekly. The strategy is intended to boost productivity and accelerate development cycles, but incidents like this underscore the risks of deploying autonomous systems in complex environments.
AWS is not just another division. It generated about 57% of Amazon’s operating profit in 2025, meaning reliability issues can have outsized financial implications even if customer impact is minimal.
Big Tech teams up to prevent the next outage
Following a separate large-scale disruption earlier in the year, AWS partnered with Google $GOOGL ( ▼ 2.08% ) on efforts to improve network resilience and prevent cascading failures across cloud infrastructure. The collaboration reflects how critical uptime has become in a world increasingly dependent on cloud computing.
The broader takeaway is that AI is moving from assisting humans to acting on their behalf. As companies push toward autonomous development workflows, the line between productivity gains and systemic risk is becoming one of the most important issues in modern tech operations.