3d render of many Traffic Cones on Orange Background

Featured image by Hector Roqueta Rivero/Getty Images

Exclusive

General Catalyst-led round values young AI safety startup Haize Labs at $100M

Founded less than one year ago by three recent Harvard graduates, Haize Labs is now valued at $100 million.

Haize Labs, an AI safety startup founded less than one year ago, has been valued at $100 million post-money in a hotly competitive round led by General Catalyst, according to two people familiar with the deal.

The startup has developed several proofs of concept for its service that stress tests large language models to eliminate vulnerabilities. AI safety evangelist Anthropic is one of its customers, according to VentureBeat.

Users have exploited weaknesses in the safety guardrails of generative AI leaders to generate harmful material, such as violent or racist content. Haize Labs is building tools to detect these loopholes so they can be closed before the public discovers them.

The nine-figure valuation for such a young startup is the latest example of how robust VC demand for promising AI deals is driving up prices in the current market. The exact amount of venture capital raised by Haize Labs could not be learned. Haize Labs and General Catalyst did not respond to requests for comment.

General Catalyst’s AI investments include co-leading the Series A and Series B for LLM startup Mistral AI, currently valued at roughly $6.5 billion, and a blockbuster $50 million seed round for AI-powered healthtech Hippocratic AI in 2023.

Business Insider previously reported that Haize Labs had received at least four term sheets with valuations ranging from $30 million to $100 million, including one offer from Coatue.

Based in New York City, Haize Labs was founded seven months ago. Its founding team comprises 22-year-old CEO Leonard Tang alongside Richard Liu and Steve Li, who all met as undergraduates at Harvard. Both Tang and Li were undergraduate researchers at the Berkeley Artificial Intelligence Research Lab.

Tang’s co-authored research paper on neural networks solving university-level mathematics problems, published in 2021 in the Proceedings of the National Academy of Sciences, has been cited over 100 times since.

Tang, who previously completed internships at Nvidia, Snap, Amazon and elsewhere, dropped out of his first year of a Stanford Ph.D. program to build Haize Labs.

Featured image by Hector Roqueta Rivero/Getty Images

Learn about our editorial standards.