Israeli cybersecurity startup Hirundo has raised $8 million in seed funding to eliminate AI hallucinations and bias and help organizations deploy safer AI models.
The investment round was led by Maverick Ventures Israel, with additional support from AI.FUND, Alpha Intelligence Capital, Plug and Play Tech Center, SuperSeed, and Tachles VC.
Founded in 2023, Hirundo has created technology that enables trained AI models to unlearn unwanted data and bad behavior, thus eliminating inaccuracies, biases, and sensitive information leakage, and reducing the risk of prompt injections.
The company’s Machine Unlearning platform, which also tackles jailbreaks, aims to address issues in the AI core without the need to retrain the model, while also boosting the accuracy of non-generative applications.
According to the company, it can remove bad behaviors and inaccuracies from AI models without impeding their capabilities, helping organizations save time and money while improving the models’ reliability and performance.
Hirundo identifies how an undesired behavior manifests in the model, outlines behavioral traits, and directs the model away from them. The solution can be used with open source models, but the company aims to make it work on gated models as well.
The platform supports both generative and non-generative systems and is used by organizations across the finance, government, healthcare, and other sectors.
“Hirundo’s solution operates like a form of AI model ‘neurosurgery’, pinpointing where in a model’s billions of parameters hallucinations originate or toxic knowledge encoded, and precisely removing it. We ensure data is reliably deleted, model accuracy is assured, and the process is scalable and repeatable,” Hirundo co-founder and CEO Ben Luria said.
Related: Horizon3.ai Raises $100 Million in Series D Funding
Related: Swimlane Raises $45 Million for Security Automation Platform
Related: Compyl Raises $12 Million for GRC Platform
Related: Maze Banks $25M to Tackle Cloud Security With AI Agents
