SAFEGUARD YOUR GENERATIVE AI WORKLOADS FROM PROMPT INJECTIONS

Summary from the TLDR InfoSec newsletter

The AWS Security team provides an overview of prompt injection techniques and an overview of defenses available in AWS. AWS suggests using Amazon Bedrock Guardrails for prompt injection defenses and input and output filtering as well as integrating a WAF and secure prompt engineering practices. It also recommends ensuring proper logging and that IAM roles associated with GenAI apps not have excessive permissions.