EquiStamp is a Public Benefit Corporation that provides evaluation implementation, data annotation, red/blue teaming, and operational support so AI safety researchers can focus on research rather than logistics.
EquiStamp is a Public Benefit Corporation that provides evaluation implementation, data annotation, red/blue teaming, and operational support so AI safety researchers can focus on research rather than logistics.
People
Updated 05/18/26CEO
Director
Research engineer
Research Engineer
Research engineer
Funding Details
Updated 05/18/26- Annual Budget
- -
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- $400,000
Org Details
Updated 05/18/26EquiStamp is a Public Benefit Corporation incorporated in Delaware that provides research operations and evaluation infrastructure for AI safety organizations. The company grew out of Rob Miles' AI safety Discord community in late 2023, co-founded by Chris Canal (CEO) and Daniel O'Connell (CTO), with Rob Miles serving as AI Advisor. The company's core thesis is that AI safety research organizations spend too much time on operational and engineering tasks that distract from high-value research. EquiStamp serves as a task-based labor pool where a core team of researchers and engineers, supplemented by a broader specialist network, implement evaluations, run baselines, annotate data, and handle administrative overhead on a flexible schedule. EquiStamp has built evaluation infrastructure for several of the most prominent AI safety organizations. For METR, they built HCAST and RE-Bench. For the UK AI Safety Institute, they implemented Control Arena. For Redwood Research, they built Linux Bench. They also built CVE Bench for Daniel Kang's lab at UIUC. These projects required deploying specialized engineering teams on short notice to meet publication, policy, and contract deadlines. The company uses a task-based work model where contributors work as much or as little as they want, as long as tasks are completed. This allows EquiStamp to surge capacity rapidly when clients need it and scale down between projects. The organization is structured to be self-sustaining, generating revenue from AI safety clients rather than relying on grant funding. A 2024-2025 Manifund raise of $400K in operating capital (structured as a SAFE investment) addressed cash-flow timing constraints rather than operational losses, as clients pay on Net 30 or longer terms while EquiStamp pays staff weekly. In addition to contract work, EquiStamp operates a platform at equistamp.com where users can run evaluations, generate reports, and manage AI alert subscriptions. The platform supports API access, TypeScript clients, and GitHub Actions integration.
Theory of Change
Updated 05/18/26EquiStamp believes that a major bottleneck in AI safety research is operational and engineering capacity: researchers spend significant time on implementation tasks that could be done by skilled but less specialized engineers. By providing a reliable, mission-aligned labor pool, EquiStamp allows researchers to concentrate on the highest-value intellectual work. In the near term, this accelerates the production of evaluations and benchmarks that are used to understand and constrain the capabilities of frontier AI systems. The company's legal structure as a PBC ensures that even as it scales commercially, its incentives remain aligned with advancing safety rather than maximizing revenue. By making evaluation infrastructure cheaper and faster to build, EquiStamp aims to reduce the cost of safety work across the ecosystem.
Grants Received– no grants recorded
Updated 05/18/26Projects
Updated 05/18/26
Benchmark of 189 machine-learning, cybersecurity, software engineering, and reasoning tasks with over 1,500 hours of human baselines, used to evaluate autonomous AI agents and to which EquiStamp-affiliated researchers contributed alongside METR.
Discussion
No comments yet. Be the first to share your thoughts.