Hirundo, a Tel Aviv-based leader in machine unlearning, has secured $8 million in seed funding to address critical AI shortcomings. Its innovative technology enables enterprises to remove biases, hallucinations, and malicious data from AI models, ensuring reliable and secure AI deployments across industries.
Hirundo raises $8M led by Maverick Ventures Israel.
Cuts AI hallucinations by 55% and biases by 70%.
Enables machine unlearning without model retraining.
Compatible with models like Llama and DeepSeek-R1.
Serves finance, healthcare, and defense sectors.
Supports scalable, mission-critical AI applications.
As generative AI becomes central to enterprise operations, issues like hallucinations—where AI generates false or misleading outputs—and biases in training data create significant risks. Hirundo’s machine unlearning technology eliminates these flaws directly within trained models, bypassing the need for expensive retraining. Unlike guardrails or fine-tuning, which only mask problems, Hirundo’s solution corrects the model’s core, enhancing reliability and reducing legal and operational risks.
Hirundo’s approach pinpoints and removes problematic behaviors, including adversarial manipulations, from AI models. "Broader adoption of AI is limited by hallucinations and undesired behaviors which make models too risky to deploy in enterprise-level applications. With Hirundo, models can be remediated instantly at their core, working towards fairer and more accurate outputs," said Ben Luria, CEO & Co-Founder of Hirundo. The technology has reduced hallucinations by 55% and biases by 70% in models like Llama and DeepSeek-R1.
Hirundo’s platform is designed for critical AI applications in finance, healthcare, and defense, supporting both generative models and non-generative systems like those used in NLP and computer vision. "Without removing hallucinations or biased intelligence from AI, we end up distorting outcomes and encouraging mistrust - for organizations and enterprises this means legal, financial and real-world consequences," said Yaron Carni, Founder at Maverick Ventures Israel. The solution is currently in pilot with multinational corporations and government agencies.
Hirundo’s technology works with open-source models like Mistral and Llama, with plans to support gated models like ChatGPT. "Hirundo’s technology is groundbreaking in addressing the hallucination and bias problems that plague even the most advanced models today," said Mads Jensen, Managing Partner of SuperSeed. The funding will enhance the platform, driving broader adoption of trustworthy AI.
Hirundo’s machine unlearning technology is revolutionizing AI reliability. By efficiently addressing model flaws, it enables enterprises worldwide to deploy AI with confidence and precision.
Founded in 2023 by serial entrepreneur and Rhodes Scholar, Ben Luria alongside award-winning researcher Michael Leybovich and Emer. Professor Oded Shmueli, former Dean of Computer Science at Israel's Technion, Hirundo is the first startup in the world to offer an unlearning solution, pioneering the concept of "making AI forget."
Hirundo's technology enables AI models to remove unwanted data or behaviors they have previously learned, ensuring that biases, sensitive information, or inaccuracies can be effectively remediated. By leveraging patent-pending technologies, Hirundo is dedicated to making AI more safe, trustworthy and accurate. Its solutions empower enterprises to deploy AI with confidence, ensuring responsible AI adoption while providing mission-critical accuracy.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones—and thrive on the tech pulse.