Product
This brief outlines a detection strategy for identifying unusually large prompts sent to AWS Bedrock Claude models, which may indicate prompt injection attacks, data exfiltration attempts, or abuse of the AI service, warranting investigation by detection engineers.