Shypd.aiThe largest open index of AI tools
Hirundo is a machine unlearning platform that removes unwanted data and behaviors from AI models. It helps optimize, debug, and ensure compliance for AI systems by surgically removing risky knowledge without retraining.
Trending
About
Hirundo is a machine unlearning platform designed to address critical risks in AI models, such as jailbreak vulnerabilities, hallucinations, bias, toxic outputs, and memorized PII. It allows users to surgically remove unwanted data and behaviors from trained LLMs at the model level, eliminating the need for lengthy retraining processes. The platform offers built-in evaluations and red-team testing to pinpoint risky behavior, then identifies and modifies specific parameters responsible for the unwanted knowledge. This results in a fixed, higher-performing model in hours, significantly reducing jailbreaks by up to 85%, biases by 70%, and achieving 100% PII removal without impacting other data or functionality. Hirundo supports both LLM unlearning and data QA for non-generative AI, ensuring models are hardened before deployment and continuously improved in production.
Capabilities
Pricing & Plans
Likely Not Free
Not publicly disclosed. Check hirundo.io for current pricing.
FAQs
Trending
