Meta on Tuesday revealed LlamaFirewall, an open-source framework intended to safeguard artificial intelligence (AI) systems from new cybersecurity threats such as prompt manipulation, jailbreaks, and insecure coding practices, among others.
The framework, as stated by the company, features three protective measures, which include PromptGuard 2, Agent Alignment Verifications, and CodeShield.
PromptGuard 2 aims to identify direct