Responding to fears of an imminent Soviet nuclear attack, in 1951 Pres. Harry Truman set up a national system enabling the president to quickly notify the public of an impending national security threat via a cross-country relay chain of AM radio stations. It used characteristic blaring warning tones and became a precursor of the Emergency Alert System still in use today. “There are certain stations across every market that listen for those tones and then retransmit the alert to other stations in their market,” says John Lawson, an emergency alert expert who has advised the Federal Emergency Management Agency (FEMA) on its modern warning systems.

A 2006 bill passed under Pres. George W. Bush augmented that capability with wireless text message alerts that could theoretically be sent to every cell phone across the country in the event of a national emergency like a nuclear attack. But surprisingly this new system has never even been tested—so Pres. Donald Trump plans to send an inaugural test alert on October 3. (It had initially been slated for September 20, but FEMA says the plan was delayed by the need to focus on Hurricane Florence.)

This week Lawson spoke with Scientific American about how the new alert system may play out. Lawson serves as executive director of a private sector technology collaboration group called the Advanced Warning and Response Network (AWARN) Alliance, and spends his days thinking about how to best prepare communities for unthinkable disasters. AWARN aims to create alerting systems that could still function if the cellular network jams or the power grid goes down.

[An edited transcript of the interview follows]

How would this presidential text message alert system work?

It’s my understanding that the code to send an emergency alert and the instructions would travel with the president in the “football”—the suitcase with the nuclear launch codes. [Asked for the agency’s perspective on the alert system, a FEMA spokesperson said she would not immediately be able to provide details or make an expert available for an interview.]

Other than a nuclear attack, what circumstances would make it appropriate for the president to send a text versus, say, one coming from the head of FEMA?

The president could theoretically use it for any other type of alert as decided by the president.

Our current president is very active on Twitter, and I think one concern is that—without some sort of filter system in place—the president could send one of these national alerts and incite panic, or even an international incident of some kind.

To your question about potential abuse, yeah, it’s possible. It does come down to a trust of civilian authority, of the commander in chief.

Let’s go back to the cold war context, which is relevant here. You have to trust somebody. Let’s say there is a massive cyber attack that takes out the grid. Ironically the old emergency alert system, because it uses AM radio and the daisy chain, would probably survive that. Whereas the new system that depends on internet distribution probably wouldn’t—it is good we have both alert systems.

So is it a good thing FEMA is setting up this presidential alert system now and testing it, even if it probably wouldn’t work in that scenario without the grid?

Absolutely. I would say it’s about time.

Who would be involved in helping to draft this kind of presidential text alert, and also handing the president the necessary instruction and whatnot to “press send”? I’m assuming he wouldn’t just be sending this from his phone like a tweet?

I’m pretty sure a lot of that is classified.

What’s the legislative background leading to this national emergency alert system?

The Wireless Emergency Alert, or WEA, system was created by the WARN Act passed by Congress in 2006 [WARN stands for Warning, Alert and Response Network]. It was designed to motivate the wireless carriers to provide a text message system for emergency alerting, and there was a procedure—a rulemaking—at the Federal Communications Commission. There was an advisory subcommittee, and I was on that, to help establish the rules that led to the present system.

WEA provides for any authorized originator, often county governments, to use that system for imminent threat emergencies. And it also allows people, users, to opt out of these alerts—except for an alert from the president of the United States. So these alerts are sent by a county government and go over the internet to an aggregator, a server managed by FEMA, to authenticate them. They are then redistributed to local wireless carriers that retransmit them to cell towers.

“Authenticate” makes it sound kind of vague and mysterious. You mean there is a person who double-checks that these are valid warnings, and has to take action? Or is it more like a CAPTCHA system (which thwarts internet spam), that just makes sure it’s a real signal?

I’m pretty sure no human touches it. A CAPTCHA would be a good analogy. It’s some sophisticated system of authentication that involves codes and passwords and other software that would detect what is legitimate and what’s not.

Okay, well what would rise to the level of needing a presidential alert beyond a nuclear threat? It seems to me that most emergencies are regional—like a flood, earthquake or even a sinkhole.

It’s always something horrible that we haven’t imagined. I can see the system being used post-attack.

If this presidential wireless alert system was created under George W. Bush, why are we only testing it now?

I think there have been many alerting failures recently, so that momentum is picking up to do something and improve our system. Hawaii [where an alert system erroneously sent text-message warnings of an incoming ballistic missile in January] was certainly a catalyzing moment. Post-Hawaii there has been a heightened awareness that we need to up our game in terms of alerting. And it wasn’t just that instance. Last year at about this time there were massive wildfires in Sonoma and the North Bay region [in California], and people died—and the wireless alert system was not used to warn people. The Sonoma emergency manager had reasons not to send them [warning texts]: One was because the number of characters [90] is so limited that you can’t really instruct people very specifically about what action they should take. If you say, “evacuate immediately,” well, in what direction? The alerts aren’t geotargeted, so you will warn people who aren’t affected—and they could then clog the roads so people can’t get emergency vehicles in.

Can people really not block these presidential alerts? Or is it like any other cellular phone number—say from an ex you don’t want to hear from—and you can just block that number?

They are not actually sending a call through the exchange to your phone, so you cannot block it. It’s actually a broadcast alert—and the WARN Act of 2006 that set up what became this wireless alert system specifically allows people to opt out of everything but the presidential alert.