Build a parallel, community-rooted AI research institution

Stand up an independent research institute outside the corporate AI labs and the large foundation-policy think tanks, structured so the research agenda is set by the communities most affected by AI rather than by funders or executives. The form is small, distributed, and explicitly counter-institutional: a thin permanent staff, a flexible network of community-embedded researchers, fiscal sponsorship or non-profit incorporation, philanthropic but not Big-Tech funding, and a publication discipline that prioritises affected-community readers alongside academic ones.

An actor chooses this strategy when the existing research apparatus — corporate labs and policy think tanks — is structurally incapable of producing the kind of knowledge the movement needs, because its funding and career incentives point elsewhere. A parallel institute solves the agenda-setting problem at the source: researchers are paid to ask the questions affected communities raise rather than the questions a sponsor will fund.

It trades off scale and durability for independence. Operating outside the dominant funding stream means perpetual fundraising pressure, fragile career paths for the researchers it employs, and a small footprint relative to the institutions it is countering. The strategy depends on a philanthropic ecosystem willing to subsidise non-aligned research — a contingent, revocable condition.