Amazon Rejects Claims That AI-Written Code Caused Website Outages

Amazon has pushed back against reports suggesting that artificial intelligence-generated code caused several recent outages on its retail website, saying the claims were inaccurate and based on misinterpretations of internal incidents.

Through a blog post, the company addressed earlier reporting by the Financial Times that linked AI tools to disruptions affecting Amazon’s online store. Amazon said the report incorrectly suggested that AI-written code was responsible for multiple outages.

According to Amazon, only one of the incidents involved AI-assisted tools, and even in that case the disruption was not caused by AI code. Instead, the company said the issue resulted from human error within an engineering team, which allowed the mistake to have a broader impact than intended.

The company also clarified that the outages discussed were separate operational issues that occurred over the course of a week, varying in duration and impact. Teams reviewed them during a regular internal operations meeting where engineers analyse service incidents and identify ways to improve system reliability.

One incident involving AI tools occurred when an engineer followed incorrect guidance generated by an AI assistant that relied on outdated information from an internal wiki, Amazon said. The company added that it has since updated internal guidance to prevent similar issues.

Amazon also rejected suggestions that its cloud division, Amazon Web Services, was involved in the outages, stating that the disruptions were limited to retail store infrastructure and did not affect AWS services.

"Some reports also suggested AWS services were involved or that Amazon introduced new approval requirements for engineers working with AI tools. That is false. The incidents discussed were limited to Amazon’s retail store infrastructure and did not involve AWS," the company said.