Following the major outage owing to a fault in its internal AI coding assistant, Amazon has now announced a temporary 90-day “code safety reset” across its critical engineering systems. A series of recent outages disrupted customer orders and exposed vulnerabilities that were eventually linked to the aggressive adoption of AI-assisted coding tools, something which Amazon now wants to fix.
The move, which has been detailed in internal documents and communications (originally reported by Business Insider) from Dave Treadwell, Amazon’s SVP of e-commerce services, aims to introduce “controlled friction” to code changes in high-impact areas while the company works on longer-term safeguards. It targets approximately 335 Tier-1 systems – services that directly affect customers and have experienced multiple order-impacting incidents since last year.
Amazon goes for a 90-day code safety reset
The damage control comes after incidents from early March 2026 highlighted reliability challenges emerging from non-deterministic AI outputs. On March 2, an outage, partially attributed to Amazon’s AI coding assistant Q, resulted in around 1.6 million website errors and nearly 120,000 lost orders globally, with customers seeing incorrect delivery times at checkout.
Just three days later, on March 5, a separate production change, deployed without proper documentation or approval, caused a sharp 99% drop in orders across North American marketplaces, leading to an estimated 6.3 million lost transactions in a single day.
In an internal note, Treadwell stated that the events were part of a “trend of incidents” since the third quarter of 2025, most of which are driven by faster code deployment created by AI tools, which lacked sufficient review and safeguards.
“We are implementing temporary safety practices which will introduce controlled friction to changes in the most important parts of the Retail experience,” Treadwell wrote. “In parallel, we will invest in more durable solutions, including both deterministic and agentic safeguards,” he added.
Amazon’s key measures under the 90-day reset
Under the new guidelines, engineers working on Tier-1 systems must now:
– Secure reviews from two individuals before deploying critical code changes (a requirement that had lapsed in some teams).
– Use Amazon’s internal Modeled Change Management tool for all production changes to ensure proper documentation and approval.
– Rely on automated coding systems that strictly follow central reliability engineering standards.
– Undergo enhanced reliability testing to verify stability.
Amazon’s Directors and VP-level leaders responsible for these systems have been directed to conduct top-down audits of code-writing, approval, and deployment processes within their organisations.
Amazon has clarified that the policy does not require a senior engineer’s sign-off, specifically for junior or mid-level engineers, on every AI-assisted change. However, oversight on these processes has been tightened overall.