The Information cites sources saying OpenAI CEO Sam Altman last week outlined three safety paths as the Trump administration prepares its first superintelligence oversight framework: an IAEA-style global body inspecting systems and compute; a “Demis Plan” merging major labs under a CERN-like entity