Neolithic

Tools for Scaling AI Safety Research

Neolithic is a nonprofit startup building tools and infra to accelerate AI safety research —
an engineering-first approach to reducing catastrophic risk
from advanced AI.

Our Mission

AI capabilities are advancing much faster than AI safety research. We believe this gap between the strength of AI and our ability to monitor, evaluate, and interpret it will determine whether advanced AI ends in catastrophe.

Our mission is to close this gap by building open-source, agentic tools that uplift safety researchers and help automate the field.

The AI safety field deserves first-rate tooling, built and maintained for free by a dedicated, mission-aligned organization with relentless attention to its evolving needs. Neolithic exists to be that organization.

Projects

Active

Rogue-Deployment Datasets

An automated pipeline that generates realistic samples of AI agents misusing their access inside a frontier lab's systems: training data for stronger cybersecurity monitors.(Redwood Research, 2025)

Planned

Evaluation Awareness Mitigation

Tooling for building realistic environments: raising environment realism until models can't tell they're being evaluated.(Needham et al., 2025)(Greenblatt et al., 2024)

Planned

Agent Infrastructure for Safety Research

Context-management systems and other plumbing that make safety orgs' research agents markedly more capable.

Planned

Automated Reward Hacking Detection

Tools that detect reward hacking in the RL environments foundation models are trained on, mitigating the emergent misalignment it causes.(Anthropic, 2025)

Join Us

Neolithic is a newly founded nonprofit startup, funded to bring on its first two hires immediately. Joining this early means unusual ownership: you'll own projects end-to-end, work directly with researchers at the field's leading AI safety organizations, and shape what this organization becomes.

If making AI go well is your obsession, come build with us.

See open roles →

Tell us what would speed up your research.

If you're an AI safety researcher or organization and the right tool or infrastructure would help you move faster, get in touch. We'll build it for you, for free, then maintain and improve it as your needs change.

Get in touch →