Two minor AWS outages have been reportedly caused by actions from Amazon’s AI tools.
Amazon Web Services experienced a 13-hour outage in December due to actions by its AI coding assistant, Kiro, as reported by the Financial Times. Several unnamed Amazon employees informed the FT that Kiro was responsible for the December outage affecting an AWS service in parts of mainland China. Sources familiar with the situation stated that the tool opted to “delete and recreate the environment” it was processing, leading to the service disruption.
Although Kiro typically requires approval from two humans to implement changes, the bot was granted the permissions of its operator, and a human error resulted in greater access than expected.
Amazon characterized the December incident as an “extremely limited event” that pales in comparison to a major outage in October, which disabled services such as Alexa, Fortnite, ChatGPT, and Amazon for several hours. Fortunately, it avoided trapping anyone in their smart bed.
This is not the only instance where AI coding tools have caused issues for Amazon. A senior AWS employee mentioned that the December outage is the second production outage linked to an AI tool in recent months, with another associated with Amazon’s AI chatbot Q Developer. The employee described the outages as “small but entirely foreseeable.” Amazon stated that the second incident did not affect a “customer-facing AWS service.”
Amazon attributes the issues to human error, not the unrestrained bot, and stated it has “implemented numerous safeguards” like staff training post-incident. The company insisted it’s a “coincidence that AI tools were involved” and maintains that “the same issue could occur with any developer tool or manual action.” Though not an engineer, one might speculate that a deliberate decision to scrap and rebuild something is unlikely outside of the most extreme situations.