Amazon Launches 90-Day Code Reset After AI Coding Outages Hit Orders

Amazon has initiated a temporary 90-day “code safety reset” across sections of its engineering organisation after a series of software failures disrupted its online retail operations and caused thousands of customer orders to be delayed or lost.

According to a report by Business Insider, the move comes after multiple technical incidents in recent months, including at least one system outage associated with the company’s internal AI-assisted coding tool, Amazon Q.

Senior Amazon executives reportedly informed employees that the company had observed a troubling pattern of software issues since late 2025. These incidents triggered concerns about how code is written, reviewed, and deployed across some of the company’s most critical infrastructure systems.

As a result, Amazon has decided to temporarily tighten its development processes to improve software reliability and prevent further disruptions.

Stricter Controls for Software Development

New Rules for Engineers

Enhanced Code Reviews and Approvals

Under the newly introduced reset programme, engineers working on core infrastructure will have to follow more stringent procedures before deploying software updates.

Key measures include:

• Additional layers of code review

• Mandatory approvals before code deployment

• Detailed documentation of software modifications

The company is also adding extra safeguards designed to slow down risky deployments and ensure higher levels of reliability.

Introducing "controlled friction"

As part of these safeguards, Amazon is introducing "controlled friction" into the code-review process. The aim is to deliberately slow down changes to sensitive systems so that potential errors are identified before software is released to production environments.

This approach is often used in large-scale technology companies to balance innovation speed with operational stability.

Major Service Incidents Trigger the Reset

Outages Affecting Orders and Customers

The reset programme was triggered after multiple disruptions affected Amazon’s retail platform in recent months.

One reported incident led to more than 100,000 customer orders being lost, while another disruption affected millions of purchases across the platform.

These outages exposed vulnerabilities in the company’s software deployment systems and raised concerns about the pace of code changes in complex infrastructure environments.

Focus on Amazon’s Most Critical Systems

Protecting “Tier-1 Systems”

The safety reset specifically targets about 335 "tier-1 systems", which are the company’s most critical services.

What Are Tier-1 Systems?

Tier-1 systems are software platforms that directly impact customers and business operations, including:

• E-commerce order processing

• Payment infrastructure

• logistics and fulfilment systems

• cloud services supporting retail operations

Failures in these systems can immediately disrupt shopping experiences for millions of customers.

Because of their importance, the company is prioritising extra safeguards around updates to these systems.

Temporary Measures to Stabilise Infrastructure

Three-Month Audit and Review Period

Although Amazon has not publicly disclosed the full details of the programme, internal reports suggest the temporary restrictions will remain in place for roughly three months.

During this time, engineering teams are expected to:

• audit existing software systems

• strengthen development standards

• identify vulnerabilities in code deployment processes

The goal is to stabilise the company’s technology infrastructure and reduce the risk of outages affecting customers.

AI Coding Tools Under Scrutiny

Rapid Adoption of Generative AI

The reset programme also highlights growing concerns about the rapid adoption of AI-powered development tools within large technology companies.

Amazon has been aggressively integrating generative AI into its internal workflows, including the use of its AI coding assistant, Amazon Q.

However, some employees have raised concerns about the reliability of AI-generated code.

According to a report by The Guardian, several staff members said the rapid rollout of AI-assisted coding tools has sometimes produced flawed software that requires significant manual fixes.

Despite these issues, Amazon continues to push forward with generative AI adoption as part of its broader strategy to increase productivity and automation across engineering teams.

AI and Software Development: A Growing Industry Debate

Speed vs Stability

The challenges Amazon is facing reflect a broader debate within the technology industry. Many companies are embracing AI coding assistants to accelerate software development, but concerns remain about quality control.

Generative AI tools can help developers write code faster, automate repetitive tasks, and reduce development cycles. However, experts warn that excessive reliance on AI-generated code without rigorous human review may introduce bugs or vulnerabilities.

As technology giants race to integrate AI into their development workflows, balancing speed with reliability has become a critical challenge.

Conclusion

Amazon’s decision to implement a 90-day “code safety reset” highlights the growing complexities of managing large-scale digital infrastructure in the age of AI-assisted software development. While generative AI tools promise greater productivity for engineering teams, recent outages demonstrate that rapid adoption without robust safeguards can lead to operational risks.

By introducing stricter review processes, additional approvals, and “controlled friction” in its development pipeline, Amazon aims to stabilise its critical systems and prevent further disruptions to its e-commerce platform.

The outcome of this reset could shape how large technology companies balance innovation with reliability as AI becomes increasingly embedded in the software development lifecycle.