Amazon Web Services (AWS) experienced at least two outages in the past few months. These outages were reportedly linked internally to the use of its own AI coding tools. This led the company to a closer examination of how these assistants are being deployed across its operations. A report claims that the biggest cloud service provider has shared an internal postmortem about the “outage” of an AWS system that will allow customers to review and analyse the costs of its services. After the investigation, Amazon has concluded that it was a “coincidence that AI tools were involved” and noted that “the same issue could occur with any developer tool or manual action.”In a statement to the Financial Times (FT), the company characterised both incidents as user-related mistakes, claiming that “in both instances, this was user error, not AI error,” and added that internal analysis found no evidence suggesting errors were more frequent when AI tools were used compared with other development methods.According to the FT report, in one instance, a 13-hour disruption in December 2025 occurred after engineers allowed Kiro, Amazon’s agentic AI coding tool, to make changes to a system, following which the tool determined that the best course of action was to “delete and recreate the environment.”The company also noted that the December 2025 incident was an “extremely limited event” that affected only a single service in certain parts of mainland China. Amazon also stated that the second incident did not impact a “customer-facing AWS service.”However, it is important to note that neither disruption reached the scale of a 15-hour AWS outage in October 2025, which knocked multiple customers’ apps and websites offline, including OpenAI’s ChatGPT.Amazon also claimed that it was seeing continued customer adoption of Kiro and that it aims to benefit both customers and employees from efficiency improvements.“Following the December incident, AWS implemented numerous safeguards,” the company said, adding that these measures include mandatory peer review processes and additional staff training.

What Amazon employees said about the AWS outages

Multiple Amazon employees told the Financial Times that this was the second time in recent months that one of the company’s AI tools had been at the centre of a service disruption. “We’ve already seen at least two production outages [in the past few months]. The engineers let the AI [agent] resolve an issue without intervention. The outages were small but entirely foreseeable,” a senior AWS employee told FT.Employees said the company’s AI tools were treated as an extension of an operator and given the same permissions as the operator. In both cases, the engineers involved did not require a second person’s approval before making changes, as would normally be the case. However, Amazon defended the Kiro tool, saying it “requests authorisation before taking any action” by default. The company also added that the engineer involved in the December 2025 incident had “broader permissions than expected — a user access control issue, not an AI autonomy issue.”In July, AWS launched the Kiro coding assistant, which the company described as a step beyond “vibe coding,” enabling users to quickly build apps without writing code by following a defined set of specifications. The company had previously relied on Amazon Q Developer, an AI-enabled chatbot, to help engineers write code. Three employees said this tool was involved in the earlier outage.Some Amazon employees said they remained sceptical of AI tools for the bulk of their work, citing the risk of error. They added that the company had set a target of 80% of developers using AI for coding tasks at least once a week and was closely tracking adoption.AWS, which accounts for 60% of Amazon’s operating profits, is working to build and deploy AI tools, including agents capable of taking actions independently based on human instructions. Like other large technology companies, it is seeking to sell this technology to outside customers. The two incidents highlight the risk that AI tools can behave in unintended ways, leading to service disruptions.