Roon Says AI Labs Should Fund Alignment NGOs

Pseudonymous OpenAI staff member @tszzl posted that frontier model companies freeload off alignment NGOs including METR, Redwood, Apollo, and Goodfire. He stated the labs offload sizeable alignment externalities onto these groups and should inject billions into organizations they support. In a visible reply, researcher Seán Ó hÉigeartaigh added that UK taxpayers also fund such work through the AI Safety Institute. The posts discuss responsibility for covering alignment costs tied to model training.