Comment by Nick Vincent

In this post, I want to lay out a relatively concrete proposal for a “data dividend” that I believe has the potential to navigate the tension between backward-looking redistribution or dividend-style policy options and forward-looking data markets. Specifically: I propose that we design and implement a presumptive commons-rent tax (PCRT) on operating profits attributable to frontier AI systems which can be credited away (possibly to zero) by providing independent evaluators with data attribution receipts arguing that specific frontier capabilities derive from either private market data or commons data which was used with some kind of reciprocity. Tax proceeds would be split between funding new public goods, making additional reciprocity and sustainability payments to data commons themselves (e.g. funding new projects to help the Internet react to new forms of agentic commerce and activity, spam and scams, etc.; paying to help maintain core Internet infrastructure; supporting peer-production projects, etc.), and making payments to individuals (if viable).