Amazon has announced a 90-day “code safety reset” across its engineering systems following a series of outages linked to the company’s increasing use of AI-assisted coding tools. The move aims to stabilise critical digital operations after recent disruptions affected customer orders across its global e-commerce platforms.
According to internal communications cited in reports, Dave Treadwell, senior vice president of e-commerce services at Amazon, told employees that the company had observed a rise in operational incidents since the third quarter of 2025. The temporary measures will introduce stricter controls and approval processes for changes made to key systems that directly impact customers.
The new policy will apply to around 335 “Tier-1 systems”, which are considered the most critical infrastructure within Amazon’s retail ecosystem. Engineers will now face tighter oversight when deploying code changes, including mandatory documentation and multiple review stages before updates can be implemented.
Also Read: Trump Trade War Returns: US Probes 16 Countries Over Industrial Capacity and Forced Labour
The disruptions were partly linked to Amazon Q, the company’s AI coding assistant designed to help engineers generate and deploy code faster. While the tool has accelerated software development, the non-deterministic nature of AI systems—meaning they may produce slightly different outputs for the same prompt—has complicated traditional software review processes that require precise and consistent results.
As part of the reset, Amazon engineers will now be required to obtain approval from at least two reviewers before implementing code changes. They must also use internal documentation and approval tools alongside automated coding systems that strictly follow the company’s central reliability engineering guidelines.
The decision follows a major disruption on March 2, when an error linked partly to the AI coding tool caused incorrect delivery time estimates at checkout. The issue reportedly generated about 1.6 million errors and led to roughly 120,000 lost orders across Amazon’s marketplaces. Another outage on March 5 caused an estimated 6.3 million lost orders after a production change was deployed without following the company’s formal documentation and approval process.
During the 90-day reset period, Amazon said it will also work on long-term improvements by combining AI-driven “agentic” systems with more predictable rules-based “deterministic” safeguards to ensure greater reliability in its software infrastructure.
Also Read: Mumbai's Vidhan Bhavan Evacuated after Bomb Threat, Police Declare Hoax After Full Inspection