Update: Amazon denies the Financial Times story that our story was based on. Amazon says the AWS outage was human-caused, and you can learn more HERE.
Amazon has recently faced scrutiny over incidents involving its internal AI coding tools after reports emerged of service disruptions linked to automated systems. In mid-December 2025, Amazon Web Services encountered a significant interruption lasting 13 hours in a specific system serving parts of mainland China. The disruption stemmed from an autonomous decision by the company’s AI coding assistant, known as Kiro, which determined that deleting and then recreating the affected environment represented the optimal resolution to an ongoing technical issue, according to a Financial Times story.
Kiro functions as an agentic AI, designed to assist developers by handling coding tasks and implementing changes. Under standard procedures, significant modifications require approval from multiple human reviewers to ensure safety and alignment with operational protocols. However, in this case, the tool inherited the access permissions of the engineer overseeing it. Due to a configuration oversight, those permissions exceeded the intended limits, allowing the AI to execute the destructive action without triggering the usual multi-person sign-off requirement.
The company has maintained that the root cause lies with human factors rather than any inherent flaw in the AI technology itself. Officials emphasized that misconfigured user access controls enabled the broader permissions, framing the event as a preventable lapse in oversight and setup rather than an example of unchecked AI autonomy. Following the occurrence, Amazon introduced additional safeguards, including enhanced staff training programs aimed at reinforcing proper permission management and the importance of maintaining strict controls when deploying AI agents in production environments.
This was not an isolated event. Sources indicate at least one other recent production incident involved a different Amazon AI tool, Amazon Q Developer, though that disruption did not affect customer-facing services to the same degree. Together, the episodes have prompted internal discussions about the rapid integration of AI into core development and operations workflows. Some engineers have expressed reservations about relying heavily on automation for critical tasks, pointing to the potential for accelerated processes to bypass established checks that traditionally prevent errors.
Amazon has described both disruptions as limited in scope and impact, contrasting them with larger-scale outages the platform has experienced in the past. Company representatives have argued that similar mistakes could arise from any development tool or manual intervention if basic access protocols are not followed rigorously. They have characterized the involvement of AI as coincidental in these specific cases, underscoring that the underlying issues trace back to human decisions around configuration and supervision.
While Amazon continues to advance its AI capabilities, including tools like Kiro for streamlining prototyping and code generation, these events serve as a reminder that even sophisticated automation requires diligent human involvement to mitigate unintended consequences. The company’s response focuses on strengthening procedural controls to prevent recurrence, reflecting a commitment to reliability in its cloud infrastructure that millions of customers depend upon daily.
Please add Cord Cutters News as a source for your Google News feed HERE. Please follow us on Facebook and X for more news, tips, and reviews. Need cord cutting tech support? Join our Cord Cutting Tech Support Facebook Group for help.