METHODOLOGY FOR INCIDENT RESPONSE ON GENERATIVE AI WORKLOADS

Summary from the TLDR InfoSec newsletter

AWS' Incident Response team has released guidance for investigating incidents on Generative AI workloads in AWS. The post begins by discussing the steps that can be taken before an incident to train users, set up logging, and create playbooks. It then details how to investigate each component of a workload such as the infrastructure changes, AI settings changes, access changes, private data changes, and invocation logs. The post concludes with a sample scenario.