Who gets to say no to a grant?
Research checked September 22, 2026. This newly examined April 2025 decision adds context to the already documented August 2026 Long-Term Future Fund closure. It is not a merger announced today.
In the April 30, 2025 announcement, Caleb Parikh, Oscar Howie and Zachary Robinson chose to combine EA Funds with the Centre for Effective Altruism (CEA), rather than establish separate organizations after leaving Effective Ventures. They argued that shared operations and stability justified consolidation. Fund managers would recommend grants; CEA would review them. The authors acknowledged possible disadvantages for competing projects. They also said fund donations would remain restricted, distinguishing grant money from operational support.
In his August 11, 2026 departure account, Parikh described adapting the grants he considered to CEA’s legal and reputational constraints. He regarded the tradeoff as reasonable for CEA but wrong for his own approach. His footnote says the new Transformative AI Fund still covers those kinds of grants and was improving its processes; this is not evidence of a categorical ban.
Our interpretation: the choice of institutional host can influence which proposals reach a decision, even without an explicit order to reject them. This is a reported effect on one grantmaker’s consideration, not a measured effect across all applicants. A host absorbs compliance costs and exposure; a specialist allocator may prefer more unusual bets. Their disagreement makes sense without assuming either insincerity or a single command centre.
The practical question is whose judgment survives a disagreement. Shared infrastructure may make otherwise impossible grants feasible; it may also discourage worthwhile projects that are expensive or uncomfortable to administer. Both explanations fit parts of these accounts. Neither tells us which explanation applies to a particular rejected application.
The next useful records are grant-specific review decisions, reasons and dates, any reconsideration route, and outcomes for comparable proposals under different hosts. That would distinguish compliance constraints, quality judgments, organizational self-protection and favoritism. No executed grant instrument or named contested award is established here. These records concern EA Funds and its host; they do not establish authority over METR or Anthropic.
See the earlier fund-transition account for the announced allocation routes and their limits.