About Bentham Research

Bentham Research is the world's first peer-reviewed, institute-backed evaluation and training-data platform for humanistic reasoning in frontier AI. We are establishing the emerging global standard by which AI systems will be measured, certified, and trained on humanities knowledge.

Our Vision

The Institute was designed from its founding to become a standards and certification infrastructure—the "Moody's of AI humanities reasoning"—whose ratings become the legally referenced and commercially required measure that every frontier model and enterprise AI deployment must carry.

We believe that as AI systems become our primary interface with humanistic knowledge, the need for academic rigour and expert-validated evaluation is not just a commercial opportunity, but a civilisational necessity.

The Humanities Blind Spot

The dominant paradigm for training and evaluating frontier AI models is built almost entirely on STEM disciplines. Benchmarks measure performance on domains where answers are objectively verifiable. However, the humanities—philosophy, ethics, political theory, theology, historiography, literary analysis—are systematically underrepresented.

The consequences are significant. AI systems fail most publicly in these domains: sycophantic responses to contested ethical questions, hallucinated historical facts, shallow analysis of politically complex situations, and failure to reason with intellectual honesty about moral uncertainty.

Bentham Research addresses this by providing thin, unverified, expert-light training data with high-quality, peer-reviewed content authored by verified domain experts.

The Team