Hunting

This brief identifies potential prompt injection or jailbreak attempts against AWS Bedrock Claude large language models by searching for specific phrases in user prompts, indicating attempts to subvert model behavior, extract data, or gain unauthorized access to resources.