The Financial Times quoted four people saying that in mid-December, artificial intelligence caused a system error at Amazon Web Services. According to the newspaper, the error started occurring after engineers allowed the Kiro AI tool to make some changes to a system used by third-party Amazon Web Services customers. It lasted 13 hours.
Several Amazon employees told the FT on condition of anonymity that this was the second time in recent months that one of its AI tools had caused a service outage. “Over the past few months, we have recorded at least two outages in production systems,” said a senior Amazon Web Services employee, adding that engineers allowed the AI agent to resolve the issue without human intervention and the glitches were minor.
Amazon Web Services launched programming assistant Kiro in July 2025. Kiro can observe how developers write code and can take tasks from the backlog and complete them itself. In particular, he knows how to complete work without constant human supervision.
According to the Financial Times, the previous outage was caused by another Amazon product supporting development, Amazon Q Developer (an AI-powered chatbot).
Amazon's press service stated that neither failure was the fault of artificial intelligence.
“In both cases, this was a user error, not an AI error,” the company said.
Amazon calls it a coincidence that both occur in AI-powered systems.
“The same problem can occur with any developer tool or when performing manual operations,” they emphasized.
The company said the December incident was “extremely limited” in scope and only affected one service in China, while the other service did not impact the company's customer communications service.
Some Amazon employees told the FT they remained skeptical about the usefulness of AI tools for much of their work because of the risk of errors. They added that the company has set a target of having 80% of developers use artificial intelligence for programming tasks at least once a week and is closely monitoring this.
