Amazon’s AI Coding Tools Spark Debate After Two High-Profile Outages

Amazon Web Services (AWS) is facing renewed scrutiny over the reliability of its artificial intelligence-powered coding tools after two separate incidents disrupted internal systems. The company has pushed back against claims that the outages were caused by AI failures, instead attributing them to human error and lapses in access controls.

The first incident occurred in December, when an AWS engineer using the company’s newly launched AI coding assistant, Kiro, inadvertently caused a disruption in a single service operating in parts of mainland China. Amazon described the event as an “extremely limited event,” emphasizing that it did not affect any customer-facing AWS services. However, the outage raised questions about the safety protocols surrounding AI-driven tools, especially when granted significant operational permissions.

A second incident followed, again involving an AWS engineer and the same AI tool. Amazon clarified that this disruption also did not impact any external services, but the recurrence of such events has led to internal debate about the role of AI in critical infrastructure.

These incidents stand in stark contrast to a major AWS outage in October 2025, which lasted 15 hours and knocked several high-profile services offline, including OpenAI’s ChatGPT. That outage was not linked to AI tools but rather to broader infrastructure issues, highlighting the different scales and causes of disruptions within AWS.

How AI Tools Are Integrated Into AWS Operations

Amazon’s AI coding tools, including Kiro and the earlier Amazon Q Developer, are designed to assist engineers in writing and optimizing code. These tools are treated as extensions of human operators, granted the same permissions as their users. In the cases of the recent outages, the engineers involved bypassed the usual requirement for a second person’s approval before making changes—a deviation from standard protocol.

Amazon explained that Kiro, launched in July, was intended to move beyond “vibe coding”—a colloquial term for rapid, informal application development—toward generating code based on precise specifications. By default, Kiro is designed to request authorization before executing any action. However, in the December incident, the engineer had “broader permissions than expected,” which Amazon characterized as a “user access control issue, not an AI autonomy issue.”

The earlier outage, involving Amazon Q Developer, underscored the risks associated with granting AI tools significant operational latitude. Three employees familiar with the matter confirmed that the chatbot was involved in that disruption, though Amazon has not provided detailed public commentary on the specifics.

Internal Skepticism and Adoption Pressures

Despite Amazon’s assurances, some employees remain skeptical about the utility of AI tools for core development tasks. Concerns center on the potential for errors and the challenges of integrating AI into workflows that demand high reliability. This skepticism exists alongside company mandates: Amazon has set a target for 80 percent of developers to use AI for coding tasks at least once a week, and it is closely monitoring adoption rates.

The tension between innovation and caution is palpable within AWS. While the company touts strong customer growth for Kiro and highlights the efficiency gains AI tools can deliver, the recent outages have prompted a reevaluation of how these tools are deployed and overseen.

Safeguards and Future Directions

In response to the December incident, Amazon says it has implemented numerous safeguards, including mandatory peer review and enhanced staff training. These measures aim to prevent similar occurrences and to ensure that AI tools are used responsibly within the organization.

AWS’s approach reflects a broader industry trend: as AI becomes increasingly integrated into software development, companies must balance the promise of automation with the need for robust oversight. The incidents at Amazon serve as a reminder that even the most advanced tools are only as reliable as the processes and people that govern their use.

Looking Ahead

As AWS continues to refine its AI offerings, the lessons learned from these outages will likely shape future development and deployment strategies. For now, the company remains committed to leveraging AI to drive efficiency and innovation, while also addressing the concerns raised by its own engineers and the broader tech community.

The debate over AI’s role in critical infrastructure is far from settled. As tools like Kiro become more sophisticated, the industry will need to grapple with questions of accountability, safety, and the appropriate balance between human and machine agency.

Tags: Amazon, AWS, AI coding tools, Kiro, Amazon Q Developer, outage, user error, access control, software development, automation, tech news, artificial intelligence, infrastructure, peer review, developer adoption, innovation, safety protocols, coding assistant, tech industry, viral tech news

,