Measure both sides
Study offensive and defensive capabilities together. Understand what a model can do and where it needs to improve.
San Francisco, California
Counterweight Labs builds evaluations and capability environments for frontier model companies. We bring the same research and engineering expertise to security consulting and AI integration.
What we do
Models can find vulnerabilities, investigate intrusions, and act across complex systems. Understanding their capability takes realistic tasks, reliable infrastructure, and evidence of what actually happened.
We build the environments to measure and develop those capabilities. Our public benchmarks explain the tasks and methods; our products support teams evaluating and training their models.
Organizations can engage our world-class researchers for penetration testing, security assessments, and application security consulting. Our AI consulting and integration work helps teams automate business and security processes.
Study offensive and defensive capabilities together. Understand what a model can do and where it needs to improve.
Check what changed in the system. Make results traceable to evidence, with assumptions and limits stated clearly.
Deliver systems people can operate, experiments they can reproduce, and findings they can act on.