Mercor AI Safety Fund Grants

About Mercor

Mercor's mission is to organize human intelligence to power the AI economy. We are a leading AI data company building the layer between human expertise and frontier models. Millions of domain experts on the platform are paid more than $4 million per day to train frontier AI models.

Mercor's APEX benchmark family measures AI's real-world impact on professional work. Mercor Enterprise brings this same infrastructure to Fortune 500 companies, helping them capture how their best people work and translate that expertise directly into agents.

Mercor is creating a new category of work where expertise powers AI advancement. This requires an ambitious, fast-paced, and deeply committed team. You will work alongside researchers, operators, and AI companies at the forefront of shaping systems that are redefining society.

Mercor is a profitable Series C company valued at $10 billion. We work in person five days a week from our offices in San Francisco, New York City, or London.

Mercor Safety Research Grants: $5 Million

One of the biggest challenges facing the industry is determining whether frontier AI is safe enough to deploy. A model can pass safety checks and still behave differently in production, making investment in safety research, evaluations, and verification critical.

Mercor is committing $5 million to fund safety research. Grants support:

• Researcher hours
• API credits
• Stipends for event and conference attendance
• Access to experts from Mercor's platform

This grant is separate from the Mercor Research Fellowship, which funds benchmark and economics research. You can apply to the fellowship here.

What We’re Looking For

We welcome proposals across the full range of AI safety research. We are particularly interested in:

• Misalignment: Deceptive alignment, goal misgeneralization, reward hacking, scheming, and situationally aware failure modes
• Sandbox escape: Containment failures, privilege escalation, tool misuse, and agents operating outside their intended scope
• Evaluation awareness: Models detecting that they are being tested and behaving differently under observation than in deployment
• Interpretability: Understanding what models are doing internally and whether that behavior can be made legible to a human reviewer
• Oversight and control: Scalable supervision, human-in-the-loop reliability, and failures that occur when a system is more capable than its reviewer
• Red-teaming methodology: More robust systems for uncovering novel failures

Strong proposals outside these areas are also welcome.

Why Apply

• Funding for researcher time, API credits, and event attendance
• Access, where useful to the work, to Mercor's expert network for human grading, red-teaming, and annotation, including lawyers, accountants, engineers, scientists, and clinicians
• Access to Mercor's internal evaluation infrastructure, subject to review
• Introductions to Mercor's network of researchers across frontier labs and academia

Who Should Apply

• Independent researchers, academics, PhD students, and small teams
• Applicants with a specific, well-scoped research question; the grant is built around the proposal rather than a generic research rotation
• People with a background in machine learning, computer science, statistics, or an adjacent field such as measurement, psychometrics, human-computer interaction, security, or social science
• Experience with agentic evaluation, reinforcement learning environments, adversarial machine learning, or systems security is a bonus

We expect grantees to publish a paper, open dataset, public methodology, or tool that the field can use.

How to Apply

Submit an Expression of Interest consisting of a one- or two-page document. It should include:

• A section covering your team, background, and research accomplishments
• A section describing your proposed research project
• A section outlining the project's expected outputs and impact, including directionally correct timelines and resource requirements