Amazon convened an emergency "deep dive" engineering meeting on March 10 after a string of high-severity outages hit its retail platform in a single week. The most serious incident knocked out Amazon's website and app for nearly six hours on March 5, preventing customers from checking out, viewing prices, or accessing account details. An internal document originally blamed "generative AI-assisted production changes" for the failures, though the reference to GenAI was later deleted.

The company has since introduced new safeguards requiring senior engineers to sign off on all AI-assisted code changes made by junior and mid-level developers. Amazon VP Greg Treadwell acknowledged that "best practices and safeguards" around generative AI usage have not been fully established. The incidents were described internally as having "high blast radius." Amazon had pushed employees toward its AI coding tool Kiro with an 80% weekly adoption target.

The episode highlights growing industry concerns about deploying AI-generated code in production systems without adequate human review. Elon Musk commented "proceed with caution" in response to the news.