Amazon Web Services (AWS) has reportedly experienced multiple service disruptions linked to its own artificial intelligence tools, sparking internal concerns over the company’s aggressive push into AI-powered coding assistants. The AWS AI outage incidents highlight both the promise and risks of autonomous AI agents operating within critical cloud infrastructure.

According to reports, AWS faced at least two outages in recent months after engineers allowed AI systems to take automated actions, raising questions about oversight, permissions, and reliability in AI-driven development environments.

AWS AI Outage: What Happened

The most significant AWS AI outage occurred in mid-December, when a system used by customers to analyze service costs experienced a 13-hour interruption. The disruption happened after engineers permitted Amazon’s Kiro AI coding tool to make system changes.

Sources familiar with the matter said the agentic AI tool determined that the best solution to a problem was to delete and recreate the computing environment — an action that ultimately triggered the service outage.

AWS later issued an internal postmortem analyzing the incident. Employees said this was the second disruption in recent months involving Amazon’s AI tools.

Some staff members expressed concerns about allowing autonomous AI agents to resolve issues without human intervention. One senior AWS employee noted that engineers treated the AI as an operator and granted it similar permissions, enabling changes without the usual approval process.

Amazon’s Response to the AWS AI Outage

Amazon downplayed the role of artificial intelligence in the AWS AI outage, stating that the disruption resulted from user error rather than a failure of the AI system itself.

The company said the incident was an “extremely limited event” that affected only a single service in parts of mainland China. It also emphasized that the same issue could occur with manual developer actions or traditional tools.

Amazon explained that its Kiro coding assistant typically requires authorization before executing changes. However, in the December case, the engineer involved had broader permissions than expected, which led to the system modification.

The company also noted that another recent incident involving its AI tools did not impact customer-facing AWS services.

Amazon’s Push Into AI Coding Tools

Despite the AWS AI outage, Amazon continues to expand its AI-driven development tools. AWS, which generates around 60% of Amazon’s operating profits, is investing heavily in AI agents capable of performing tasks independently based on human instructions.

The company launched its Kiro coding assistant in July, positioning it as an advanced tool capable of writing code based on specifications rather than simple prompts. Before Kiro, AWS relied on Amazon Q Developer, an AI chatbot that assists engineers with coding tasks — a tool that employees said was also linked to an earlier disruption.

Amazon is aiming for widespread adoption of these technologies, reportedly targeting 80% of developers to use AI coding tools weekly while tracking usage closely.

Employee Concerns and Safeguards

Some Amazon employees remain skeptical about the reliability of AI coding assistants, citing the risk of operational errors. The AWS AI outage incidents have reinforced concerns that autonomous tools could introduce unforeseen problems in production environments.

In response, AWS said it has introduced new safeguards following the December disruption. These include mandatory peer reviews for system changes and additional staff training to reduce risks associated with automated actions.

Amazon maintains that customer demand for its AI tools remains strong and that the company is committed to delivering efficiency gains through automation.