Hirundo, an Israeli cybersecurity startup, has secured $8 million in seed funding to develop technology that enables AI models to unlearn biases and inaccuracies without retraining. This innovation aims to enhance AI safety and reliability across various sectors like finance, healthcare, and government. #AIhallucinations #ModelBias #AI.Tansparency
Keypoints
- Hirundo’s platform removes unwanted behaviors from AI models without retraining them.
- The company’s technology addresses hallucinations, biases, and prompt injections in AI systems.
- It supports both generative and non-generative models, including open source and gated models.
- The solution acts like ‘neurosurgery,’ precisely deleting toxic knowledge from the model.
- The platform is used across sectors such as finance, government, and healthcare to improve AI accuracy and safety.
Read More: https://www.securityweek.com/hirundo-raises-8-million-to-eliminate-ais-bad-behavior/