Amazon calls meeting to investigate AI-related web outages

Amazon organised a meeting on Tuesday to investigate a “trend of incidents” related to generative AI-based coding in its ecommerce business, the Financial Times has reported.

A briefing note seen by the paper said that these incidents have been characterised by a “high blast radius” and caused by “Gen-AI assisted changes”. Employees have been using generative AI in novel ways for which the company has not yet established “best practices and safeguards”, the note said.

There are no specific incidents mentioned in the note, but the Amazon website experienced a six-hour outage on 6 March, which the company said was “related to a software code deployment”.

Business Insider additionally reported that at least one of the outages was blamed on Amazon’s own AI coding assistant, Q, which is also available as a SaaS product.

Dave Treadwell, Amazon’s senior vice-president of e-commerce services, told employees in an email that its “weekly This Week in Stores Tech” meeting would be a “deep dive into some of the issues that got us here as well as some short immediate term initiatives”, acknowledging that site availability has been a problem in recent months.

This is not the first such incident — in December, an Amazon Web Services outage was caused by another AI coding tool, Kiro, which deleted and recreated the environment after being granted certain access privileges.

Amazon has invested heavily in AI in recent years, including a €33.7 billion investment in Spanish data centres, announced 4 March. At the same time, it has cut thousands of jobs, laying off 16,000 people in January alone.