Amazon Enforces 90-Day AI Code Safety Guardrail After Costly Outages Amid Industry-Wide AI Coding Rush
Summary
Amazon enforces a 90-day safety guardrail on AI-generated code for consumer-facing systems following costly outages, requiring stricter documentation and dual approvals, as experts warn that AI code often appears production-ready before being properly tested amid an industry-wide AI coding rush.
Key Points
- Amazon is implementing a 90-day safety guardrail on AI-generated code for consumer-facing tier-1 systems after a string of outages, including a major incident in early March that took down its website and shopping app for several hours, costing the company millions in lost orders.
- Developers at Amazon are now required to document code changes more stringently and obtain approvals from two additional staff members before pushing updates, as the company also works to develop combined agentic and rules-based safeguards.
- While enterprises across the tech industry are racing to adopt AI coding tools and cutting software engineering jobs at companies like Atlassian, Oracle, and Block, experts warn that AI-generated code often appears production-ready before it has been properly pressure tested, emphasizing that human engineering judgment remains essential.