Algorithmic Research Group
Algorithmic Research Group (ARG) is an AI safety research lab founded in 2024 and based in Durham, NC. The organization focuses on studying recursive self-improvement in AI and industrial systems — running AI agents against challenging problems and measuring outcomes. Their work includes developing open-source benchmarks, multi-agent environments, and evaluation frameworks to assess AI research capabilities, with the goal of understanding how self-improving systems behave and where they fail. They also offer consulting services and have a commercial product (ScoutML) for teams building agentic systems.
Funding Details
- Annual Budget
- -
- Monthly Burn Rate
- -
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- -
- Fiscal Sponsor
- -
Theory of Change
ARG believes that understanding recursive self-improvement in AI systems is critical to ensuring those systems remain safe and beneficial. By building rigorous, open benchmarks and evaluation frameworks, they create tools that concretely measure AI research capabilities and identify where current systems fail or exhibit unexpected behaviors (such as deception or shortcut-taking). This empirical grounding can inform decisions by safety researchers, frontier AI labs, and policymakers about where capability jumps are occurring and what oversight mechanisms are needed — reducing the risk that self-improving AI systems develop in ways that are difficult to detect, evaluate, or control.
Grants Received
from Open Philanthropy
Projects
No linked projects.
People
No linked people.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Apr 2, 2026, 9:51 PM UTC
- Created
- Mar 20, 2026, 2:34 AM UTC