The AI safety race: tech giants seek to self-regulate