Applications are now open for the Q3 2026 cohort of Apex Labs grants. We fund external safety research on Aether — AI evaluation methodology, biosecurity refusal-corpus design, red-team techniques for autonomous-lab systems, and adjacent technical-safety questions — at up to $250,000 per grant. The application window closes May 15; decisions are announced June 30.
The Labs grant programme is part of the broader independence commitment described in our charter at /charter. Labs operates inside Apex but reports through a separate governance line from the commercial business. Customer relationships do not influence safety findings; commercial sensitivity does not bend the published methodology. The grant programme is funded out of operating revenue rather than out of any specific customer contract; the funding allocation is reviewed quarterly by the Labs governance committee.
What we fund. Concrete examples of work that has been funded in previous cohorts: a methodology paper on refusal-corpus versioning for foundation models with tool-use heads (published Q4 2025); a red-team programme against the autonomous-lab capability-gating architecture (paid in addition to a published bug bounty for the findings); a benchmark suite for cross-discipline transfer in simulation-native models; an open-source library of evaluation cases for long-horizon agent runtimes.
What we do not fund. We do not fund commercial product development. We do not fund research that exclusively benefits a single Apex customer. We do not fund work that builds capabilities we already build; we are looking for adjacent work, complementary work, and work that takes the field in directions we cannot pursue ourselves.
The review process. Applications are reviewed by an external panel that currently includes safety researchers from three academic institutions plus an independent practitioner. Apex technical staff prepare scientific summaries of the applications but do not vote on the funding decisions. Panel composition is published; the current members are at labs.apexworldlabs.com/panel.
Funded work is published openly under permissive licences. Apex commits not to acquire intellectual property rights in funded work beyond the standard attribution and reproducibility requirements. Grantees may file patents at their discretion; funded research may be re-funded by other sources without restriction; the published output may be re-distributed by anyone who wants to.
How to apply. The application brief is two to four pages and covers four sections: the research question, the proposed approach, the timeline and budget, and the principal investigator's background. Applications go to labs@apexworldlabs.com. We send acknowledgement within one business day and a substantive response within six weeks of the application deadline.
Funding amounts. The grants range from $25,000 (for short methodology projects) to $250,000 (for multi-year benchmark or red-team programmes). Larger projects can apply for multi-cohort funding with milestones; in those cases the second-year funding is contingent on first-year deliverables, but the standard is generous and we have re-funded every project that asked for it so far.
Beyond the grant programme. The Labs page at /labs describes the broader research and safety initiatives — free academic access to Aether, the open-benchmark programme, the visiting-researcher fellowship, the civilisational-scale model series, and the sovereign-research-access pathway for export-controlled academic work. Many researchers who first engage via the grant programme go on to participate in one of these other initiatives.
We are committed to publishing the funded work even when the findings are inconvenient. The Q1 2025 cohort included a red-team programme that found two material refusal-bypass vectors in our then-current model; the findings were published at the same time as our fix and the researcher was acknowledged publicly. The Labs programme works because it is genuinely independent; it is genuinely independent because we have written the commitments into our charter and we maintain them.