Australia's first national AusAlert test reached 94% of targeted cell towers, according to Emergency Management Minister Kristy McBain. Some people nevertheless reported receiving no alert, while others saw a notification without the expected sound. The government measured tower broadcast. The public experienced devices. [1]
That denominator difference is the story. A tower can transmit successfully while an eligible phone is outside coverage, offline, incompatible, running old software, or affected by another delivery problem. A device can receive a notice without sounding loudly enough to warn its owner. Ninety-four percent cannot describe all those outcomes.
The test ran at about 2 p.m. AEST before an intended October deployment for fires, floods, terrorism, and health emergencies. [1] It is a completed test, not proof of launch readiness or performance during a live disaster.
Towers are necessary and insufficient
Cell-broadcast systems send messages through infrastructure rather than addressing every phone as an individual text. That can improve speed and geographic targeting. It also creates several points where a successful network act may not become a successful human warning.
The government did not publish in the Guardian report the number of towers targeted. [1] Without numerator and denominator, readers cannot convert 94% into a tower count or see whether the missed towers clustered in particular regions.
Device compatibility was bounded by hardware and software. Before the test, officials expected recent iPhones and Apple Watches on specified operating systems and Android devices running Android 12 or later to receive alerts. [1] Those eligibility rules exclude some functioning devices by design.
Guardian Australia received reports involving apparently eligible phones that did not display a notification despite being switched on. It also received reports of notifications without sound. [1] Those accounts establish that failures occurred. They do not establish a national device failure rate because the sample was not representative and the total eligible-device denominator is unknown.
Receipt, sound, and comprehension are separate
An emergency warning has several stages. Infrastructure broadcasts it. A device receives it. Software displays it. Sound or vibration attracts attention. The user reads and understands it. The message arrives early enough for action. Accessibility features make it usable. Each stage can succeed or fail independently.
The test produced vivid evidence of sound in parks, a library, and a cinema. [1] Those scenes show reach in named places. They do not establish volume, duration, readability, or accessibility across devices and settings.
One user described the alert as not especially loud. Others reportedly clicked it away quickly. [1] A live warning needs a recoverable message history and clear instructions so accidental dismissal does not erase the action required.
Latency also matters. A warning that reaches nearly every device over a long interval can be less useful than a slightly smaller reach delivered before a fire front or flash flood arrives. The assigned source does not provide receipt-time distribution.
Safe phones create a different risk
Domestic and family violence advocates warned that an audible test could expose hidden safe phones. Officials advised people with hidden devices to switch them off for a 24-hour period. [1] That precaution may protect secrecy during a scheduled test, but live emergencies will not arrive with a week's publicity.
The October system therefore needs a mitigation that does not force users to choose between emergency information and personal safety. That could involve device settings, controlled opt-outs, advocate guidance, or other design changes, but the source does not establish which remedy will be adopted.
Success for this group is not simply receipt. A phone can receive and sound exactly as designed while creating danger for its owner. Population reach and safe reach are different outcomes.
No communication channel reaches everyone
McBain said no approach works 100% of the time and described AusAlert as another tool alongside existing emergency communication. [1] That is a reasonable limitation. It should not become a substitute for publishing performance data or fixing avoidable failures.
Redundancy is essential because devices fail, networks lose power, languages and disabilities differ, and people may be asleep, driving, or outside coverage. Sirens, radio, television, apps, local officials, and person-to-person warnings remain part of the system. The test should identify which groups need those backups most.
No verified X post was authorized. User reports in the Guardian establish bounded failures; they should not be inflated into claims that the entire system failed or that nobody heard it. Likewise, the government's tower figure should not be inflated into 94% of Australians reached.
The ordinary Guardian page and its Content API version are one report, not two independent witnesses. [1] [2] The API preserves the cutoff-safe publication and modification record. Citation of both is custody, not corroboration.
October needs a device denominator
The National Emergency Management Agency should publish targeted and successful tower counts, attempted and received device measures, sound rates, latency, model and operating-system splits, carrier and geographic patterns, accessibility results, and safe-phone findings. It should explain what changed after each failure.
A retest would then show whether fixes worked. Launch should remain a separate stage from testing, and live-event performance a later one. An alert used during a flood or fire will face congestion, outages, damaged towers, movement, and urgent timing absent from a scheduled national exercise.
The test succeeded in one important sense: it exposed the difference between infrastructure reporting and human receipt before the system went live. The minister can correctly say 94% of targeted towers broadcast. [1] A person holding a silent phone can correctly say no warning arrived in a useful form.
Both records belong in the evaluation. Public safety begins where the tower percentage ends.
-- MAYA CALLOWAY, New York