Search
Meta Releases LlamaFirewall, an Open-Source Security Framework for AI Agents
2025-05-09 /
Read about 0 minute
Author：小编
Meta has recently made the LlamaFirewall security framework available as open-source software, designed to offer robust system-level protection for AI agents in production environments. This framework comprises three core modules: PromptGuard 2, AlignmentCheck, and CodeShield. These modules safeguard against prompt injection attacks, ensure behavioral alignment, and prevent the generation of unsafe code, respectively. Test results indicate that LlamaFirewall not only significantly reduces the success rate of attacks but also maintains high usability for various tasks.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
