Warnings about existential AI risks should not distract governments from present-day cybersecurity harms