UK AISI's Alignment Project issues first grants; OpenAI contributes $7.5 million

The Alignment Project, a global research fund created by the UK AI Security Institute and administered by Renaissance Philanthropy, announced its first round of grants: 60 projects selected from more than 800 applications submitted by 466 institutions across 42 countries. Individual awards typically ran from £50,000 to £1 million, with some recipients also offered compute access and expert support. Total support for the fund reached £27 million, including £12 million in new funding from additional partners beyond the initiative’s founding backers.

OpenAI announced a $7.5 million contribution to the fund as part of that total, saying the money would not create a new selection process or influence which projects were chosen — it would simply expand the pool of vetted projects the existing process could support. OpenAI framed the value of independent, externally funded research as its freedom to pursue ideas that “may not align neatly with any one organisation’s roadmap,” an implicit acknowledgment that alignment research conducted inside frontier labs carries commercial and reputational pressures that outside researchers do not face.

The funded portfolio spanned mathematics, learning theory, economics and cognitive science, including work on making model reasoning more transparent, applying mechanism design to AI control problems, and making the behaviour of training optimisers more predictable — an attempt to broaden alignment research beyond the small number of labs currently building frontier models.