CTGT builds a deterministic governance layer for enterprise AI systems. Its Policy Engine sits between AI models and end users, evaluating outputs against a structured policy graph constructed from an organisation's specific data, rules, and regulations. The system enforces these policies in real time, aiming to ensure compliance and reliability in deployments where AI behaviour must be tightly controlled.
The company's technical work spans AI alignment, AI reliability, generative AI governance, and deterministic policy enforcement. In testing, the Policy Engine doubled legal reasoning accuracy from 39% to 78%, maintained 96% integrity in identity resolution tasks, and achieved up to 80% lower inference costs when paired with open-source models. CTGT targets regulated industries including finance, healthcare, and government.
CTGT was born out of research conducted at Stanford University and is backed by Y Combinator. The company focuses on making AI systems trustworthy for organisations operating under strict regulatory requirements.
