はてなブックマーク

Secure your AI from prompt to productionOpen-source LLM security trusted by 125,000+ users Adaptive red teaming that targets applications, not just models Generate customized attacks for your use case: npx promptfoo@latest redteam setupOur language models probe for specific risks in your system. Direct and indirect prompt injectionsJailbreaks tailored to your guardrailsData and PII leaksInsecure t