For years, the battle for AI safety has been fought on the grounds of accuracy. We worried about "hallucinations" – the AI making up facts or citing non-existent court cases. But as Large Language ...