Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn about LlamaFirewall, an open-source security-focused guardrail framework designed to defend against security risks associated with AI agents, in this 33-minute conference talk from The Diana Initiative 2025. Discover how large language models have evolved from simple chatbots into autonomous agents capable of performing complex tasks, and understand the new security risks these capabilities introduce that existing guardrails don't address. Explore the framework's three powerful guardrails: PromptGuard 2, a universal jailbreak detector with state-of-the-art performance; Agent Alignment Checks, an experimental chain-of-thought auditor that inspects agent reasoning for evidence of misalignment; and CodeShield, an online static analysis engine that prevents the generation of insecure or dangerous code by coding agents. Examine the easy-to-use customizable scanners that enable any developer who can write a regular expression or an LLM prompt to quickly update an agent's security guardrails. Understand how LlamaFirewall is utilized in production at Meta and learn about the open-source release that invites community collaboration in addressing new security risks introduced by AI agents. Gain insights into real-time guardrails for agentic applications that support system-level, use-case-specific security policy definition and enforcement, presented by Stephanie Ding, a software engineer at Meta specializing in security and safety for generative AI.
