Amazon Requires Senior Engineer Approval for AI Code Changes After Outages
▼ Summary
– Amazon is holding a meeting for engineers to analyze a recent trend of outages, some linked to the use of AI coding tools.
– The company noted these incidents have a wide impact and involve new AI usage where best practices are not yet fully established.
– A senior executive acknowledged that the availability of Amazon’s site and infrastructure has been poor recently.
– A specific recent outage lasted nearly six hours and was caused by an erroneous software code deployment.
– The meeting will focus on understanding the causes of these issues and implementing immediate initiatives to prevent future outages.
Amazon is implementing stricter oversight for code changes, particularly those involving generative AI tools, following a series of significant service disruptions. The company’s ecommerce division has called for a major engineering meeting to conduct a thorough analysis of recent outages, several of which have been linked to AI-assisted development. Internal documents indicate a concerning pattern of incidents with widespread impact, where the use of emerging AI coding technology was a contributing factor. The briefing notes point to a lack of fully established best practices and safeguards for these novel generative AI applications as a key issue.
Senior vice-president Dave Treadwell addressed employees in an email, acknowledging that the availability of the site and its supporting infrastructure has been subpar. He confirmed that the weekly technical meeting for the stores division would be entirely dedicated to a deep dive into the root causes of the problems. The session will also outline immediate, short-term initiatives the company hopes will prevent similar outages from occurring.
The internal note did not list the specific incidents slated for discussion. However, one major event occurred earlier this month when Amazon’s main website and mobile shopping application were inaccessible for close to six hours. The company attributed that prolonged outage to an incorrect software code deployment, which prevented customers from finalizing purchases or accessing essential features like account details and product pricing. Treadwell, who previously held a senior engineering role at Microsoft, emphasized the urgency of addressing these systemic issues to restore reliability for the massive online retail platform.
(Source: Ars Technica)
