Haize Labs
Automated red-teaming and evaluation platform for LLMs and AI agents.
Acquired — still operating
acquired by Beacon Software Sep 2026
About
Haize Labs, founded in 2024 by Leonard Tang, Steve Li, and Richard Liu, built an automated red-teaming platform for generative AI. Its suite comprised Judge (turning business rules into measurable evaluation standards), Haize (dynamic adversarial testing), Monitor (production reliability tracking), and Robustify (hardening from discovered weaknesses). The company disclosed vulnerabilities to frontier labs and reported work with OpenAI, Anthropic, AI21, and Scale AI.
It raised a $12.5 million seed round in August 2024 led by General Catalyst at a $100 million valuation. On September 17, 2026, Beacon Software announced it had acquired Haize Labs for undisclosed terms; the team became Beacon's Applied AI Research Group with co-founder Leonard Tang as VP of AI research.
Caveats: the acquisition makes the standalone product's roadmap uncertain, so treat this as a historical entry and verify current availability before depending on it. Some traction figures were company-reported.
What it can do
- Automated adversarial red-teaming
- Business-rule evaluation standards
- Production reliability monitoring
- Vulnerability disclosure workflow
- Pre-release model safety testing
Setup
Historically: sign up for the hosted platform and connect model endpoints for evaluation runs.
Prerequisites
API access to the models under test; evaluation criteria definitions.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 82; Trending score 70/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.