Holistic has unveiled an open-source library to help counter AI development risks and build fairer and more responsible systems.
The library – dubbed Holistic AI OSL – arrives at a crucial moment when organisations are increasingly deploying AI systems across sensitive domains including recruitment, healthcare, and financial services. Recent studies suggest that 65% of AI researchers and developers still consider bias a significant challenge in their work.
Holistic AI OSL tackles five primary technical risks associated with AI development through a comprehensive suite of tools. The library’s bias mitigation capabilities are particularly noteworthy, offering over 35 bias metrics across five machine learning tasks and 30 distinct mitigation strategies.
Beyond bias detection and elimination, the toolkit provides robust features for model explainability—helping developers understand and communicate how their AI systems arrive at decisions. The library also addresses system robustness, ensuring consistent performance even when confronted with adversarial attacks or data variations.
“Our new library equips organisations with tools for all AI risks, including explainability, robustness, and bias,” explains Adriano Koshiyama, Co-CEO of Holistic AI. “It supports measurement, reporting, and mitigation at every stage of the AI lifecycle, offering one of the most advanced solutions for improving quality in AI applications today.”
The library places significant emphasis on security, incorporating privacy-preserving features such as anonymisation techniques and defences against attribute inference attacks. Additionally, it helps developers balance the often-competing demands of accuracy, fairness, robustness, and security through comprehensive real-world testing scenarios.
Global insurance giant MAPFRE, which operates in nearly 40 countries and serves over 30 million customers, has already integrated the library into its AI development stack.
César Ortega, Expert Data Scientist at MAPFRE, commented: “What sets this library apart is its depth—it’s not just about identifying AI risks but actively addressing them with proven, industry-ready mitigation techniques, making it an essential part of any ethical AI development toolkit.”
The Python library is available free of charge and without licensing requirements.
(Photo by Sarah Cervantes)
See also: Anthropic launches Message Batches API for Claude
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.
Explore other upcoming enterprise technology events and webinars powered by TechForge here.
Tags: AI, algorithm, artificial intelligence, bias, coding, development, ethics, holistic ai, library, open source, open-source, privacy, programming, security