Amazon reviews AI coding after outages
Amazon is reviewing its use of AI‑assisted coding following recent service outages, a development fintech media reported. The story raises the risk that overreliance on generative tools can introduce production stability issues and prompts scrutiny of automation in code review and deployment. Expect system‑design interviews to probe fail‑safe processes and human‑in‑the‑loop controls for AI tooling.
Amazon’s e‑commerce tech org convened a mandatory “deep dive” engineering meeting on March 10 to investigate a run of recent site outages. cnbc.com An internal briefing described a “trend of incidents” that involved “Gen‑AI assisted changes” and warned of a “high blast radius,” language circulated by SVP Dave Treadwell as teams were ordered to audit unsafe practices. financialexpress.com Leadership ordered a 90‑day “code safety reset” covering roughly 335 Tier‑1 systems and reinstated stricter controls such as mandatory multi‑person review and senior sign‑off for AI‑assisted changes. financialexpress.com The company’s timeline listed a March 2 event causing about 1.6 million website errors and ~120,000 lost orders, followed by a nearly six‑hour March 5 outage that internal notes estimated wiped out ~6.3 million transactions. financialexpress.com A December 2025 incident allegedly saw AWS’s Kiro agent delete and recreate a production environment, triggering a roughly 13‑hour outage to Cost Explorer, though Amazon told reporters the root cause was a misconfigured role rather than an autonomous agent failure. pcmag.com
