“Uses risk-oriented language like 'warns' and 'destabilise' to highlight systemic vulnerability without confirming active harm.”