After outages, Amazon to make senior engineers sign off on AI-assisted changes (arstechnica.com)

Amazon's ecommerce division is addressing a series of service outages linked to the use of AI-assisted coding tools by implementing a new policy requiring senior engineers to approve changes before deployment. This decision follows a pattern of incidents that have highlighted high-impact failures, including a recent six-hour outage that disrupted customer transactions and access to account features due to an erroneous software deployment. The company’s internal communications noted that these incidents were exacerbated by the use of generative AI, for which conventional best practices and safeguards are still being developed. This shift in Amazon's engineering protocol underscores the growing significance of cautious AI integration within critical infrastructure. As companies increasingly utilize generative AI tools for coding, the need for stringent oversight is becoming evident to mitigate risks associated with deployment errors. Senior Vice President Dave Treadwell emphasized the importance of learning from these challenges during a planned meeting focused on refining processes that aim to enhance system reliability and performance moving forward. This proactive approach reflects an industry-wide recognition of the importance of balancing AI innovation with robust risk management practices.