Drawing on aviation’s trusted incident-reporting model, a former pilot argues AI safety needs confidential reporting paired with public learning, and that legitimacy, international participation and transparency must be built into SAFE from the outset.