“Framing highlights the repetition of security failures ('fourth of its kind') creating concern about systemic AI safety risks.”