Amazon Web Services (AWS) confirmed that recent service disruptions were caused by user error, not flaws in its artificial intelligence coding tools. The company stated that its AI assistants, Kiro and Amazon Q Developer, were involved in incidents in December and an earlier period, but emphasized that these were not instances of AI autonomy failures.
Employees reported that the AI tools were granted broad permissions, bypassing standard approval processes. In the December incident, an engineer had broader access than expected, a user access control issue, according to Amazon. The company launched Kiro in July, designed to write code based on specifications, moving beyond simpler "vibe coding."
Despite employee skepticism regarding AI's reliability for complex tasks, Amazon is actively promoting adoption, aiming for 80 percent of developers to use AI for coding weekly. Following the December event, AWS implemented stricter safeguards, including mandatory peer reviews and staff training, to enhance efficiency and prevent future disruptions.