Amazon Announces 90‑Day Code Safety Reset After AI‑Related Outages: What It Means for Customers and Tech

 

Amazon has announced a 90‑day code safety reset across its critical engineering systems in response to a series of recent outages that have disrupted customer experiences. This initiative comes amid rising concerns about the integration of artificial intelligence (AI) tools in software development and the reliability of automated code changes.

Amazon office building with engineers reviewing code
Amazon initiates a 90-day code safety reset to prevent AI-related outages

What Happened: AI‑Related Outages and System Disruptions

In early March and the months leading up to it, Amazon experienced several outages on its e‑commerce platform that affected customer orders and shopping activity. At least one outage on March 2 disrupted order processing, caused incorrect delivery times, and resulted in tens of thousands of lost orders, according to internal assessments.

These outages were partly linked to the company’s use of AI coding tools  including its internal coding assistant  to generate or modify production code. As companies increasingly adopt generative AI (GenAI) for coding tasks, the lack of established best practices and safeguards has become a challenge for maintaining stability in complex technical environments.

The 90‑Day Code Safety Reset Explained

To improve reliability and prevent further interruptions, Amazon has initiated a 90‑day code safety reset with stricter internal controls, including:

  • Enhanced review procedures: Engineers will now need multiple levels of approval before code changes  especially those assisted by AI  are deployed to critical systems.

  • Documentation requirements: All code changes must be comprehensively documented, allowing teams to trace and audit changes before they go live.

  • Controlled friction: The reset aims to slow down the deployment of changes in high‑impact systems to reduce the likelihood of errors.

  • Deterministic safeguards: Amazon will combine AI tools with more traditional, predictable systems to create a safer development environment.

The reset focuses on around 335 Tier‑1 systems that have a direct impact on the customer shopping experience, such as order processing, checkout systems, and delivery status updates.

Why This Matters: Reliability in the Age of AI

The growing use of AI in software development has brought both productivity improvements and new risks. Unlike traditional systems, generative AI models can produce different outputs for the same input, which makes them less predictable without human oversight.

According to internal communications cited by media, the trend of such incidents began around late 2025, driven by increasing reliance on AI tools without established controls or safeguards. The reset signals that Amazon is re‑balancing its approach  emphasizing safety over speed in critical areas of its infrastructure.

Broader Industry Context

Amazon is not alone in this transition. Across the technology sector, companies using AI for development are grappling with how best to integrate these tools while preserving system integrity. Recent reports indicate that engineers at Amazon and other tech firms are reassessing workflows to prevent AI‑generated code errors from reaching production systems.

This shift comes at a time when many companies also face workforce changes and pressure to adopt AI tools to boost efficiency, even as they navigate the risks associated with automation and reliability.

Customer Impact and What Comes Next

For shoppers, the outages may have caused delayed orders, incorrect delivery timelines, or temporary issues with browsing and checkout functions. Amazon’s reset is designed to minimize such disruptions going forward by tightening how software updates are managed.

For the tech industry, this move highlights a larger lesson: AI adoption in critical systems must be matched with robust review processes and human supervision. As Amazon continues to refine its engineering safeguards, other companies will likely watch closely and adapt similar protocols to ensure stability in their own AI‑aided operations.

Comments

Popular posts from this blog

Shyam Maheshwari SSG Shares Insights on SVB’s Collapse and the Future of Startup Banking

Shyam Maheshwari SSG: The Journey of a Global Investor from Bihar to Singapore’s Financial Hub

Gunman kills 4 before shooting himself at hospital in US