Resect AI raises $25M to cut AI hallucinations

Resect AI's approach starts from a different angle than most AI startups: rather than building a bigger model, it built its own models first and then discovered that the training techniques could be applied to existing open models including DeepSeek, Qwen and Llama. That discovery shifted the company's focus from building a better model to building a suite of tools that work within the model architecture itself to intercept misbehavior before hallucinations happen.

Chief Executive Kevin Owens said AI has prematurely been put in a position of trust, and adding labels such as "use at your own risk" flies in the face of proper governance or compliance. The company's name, "Resect," is a medical verb meaning to cut out or remove damaged tissue while leaving healthy tissue untouched, which reflects its surgical approach to fixing model errors.

The funding came from private equity investors whose names were not announced, and will go toward research and development, go-to-market initiatives and local talent hiring in the greater Seattle and Portland regions.