Artificial intelligence is already summarizing our meetings, doing our math homework and illustrating our PowerPoints.

Now, it turns out, AI also can help detect impending earthquakes.

Point in case: In late June, a 3.7 magnitude earthquake struck near the Dead Sea in Israel — and while it may have come as a shock to some of the people who felt it, it was actually detected fairly well in advance by SeismicAI, an Israeli startup based in Tel Aviv that’s developing an AI-powered earthquake warning system.

Detecting earthquakes is the holy grail for seismologists and governments as they are notoriously difficult to predict. There are roughly 55 earthquakes a day around the world, or 20,000 a year, according to the National Earthquake Information center.

While most earthquakes are small, in the last month alone, there have been 13 significant earthquakes over magnitude 5.0, in California, Russia, Utah, Guatemala and Japan.

The biggest of these at magnitude 7.1 was in Japan, and it prompted the Japanese Meteorological Agency to issue its first-ever ‘megaquake alert’ on August 8 as experts feared the original quake could have altered stresses along a larger and more dangerous fault line.

Japan is no stranger to mega quakes. In March 2011, a magnitude 9.1 earthquake off Japan’s Oshika peninsula killed 18,000 people and caused a 40 meter (130 foot) tsunami. 

Founded by a team of engineers and seismologists, Seismic.AI aims to provide faster, more accurate earthquake alerts that can save lives and minimize damage. 

Making (P) waves

What sets SeismicAI apart from traditional earthquake detection systems is its global applicability. 

Traditional systems are typically designed for specific regions and require years of local data to function effectively. SeismicAI claims its platform can be rapidly deployed anywhere in the world while maintaining high accuracy.

“We have a globally accurate system,” says Benny Sasson, the company’s CEO and cofounder. “We’ve tested our system in 15 regions worldwide and proved that it works.”

SeismicAI’s platform integrates with existing seismic sensor networks, analyzing the data in real time to distinguish earthquakes from other ground vibrations. 

When an event is detected, the system rapidly estimates the earthquake’s location, magnitude, and potential impact on surrounding areas.

“When an earthquake starts, it emits several types of seismic wave. The first one is a fast moving ‘P wave’ that you usually can’t feel. But the sensors in our system can detect it, and from that preliminary wave, we are able to estimate the probable impact of the destructive wave that will follow,” Sasson tells ISRAEL21c. 

He explained that the system can detect seismic events and issue warnings seconds to tens of seconds before destructive waves hit. 

This critical time can allow people to take cover and automated systems to shut down sensitive infrastructure — while also minimizing the risk of false positives that can cause their own forms of damage.

“The impact of a false alert can be as destructive as an earthquake itself,” Sasson said, noting the logistical issues that can be caused by triggering unnecessary emergency responses and evacuations.

A customer-oriented approach

While SeismicAI’s technology has far-reaching implications for public safety, the company operates as a private entity, serving a range of clients, rather than functioning as a public service.

That said, the 18-worker-strong company has strategically partnered with both public and private sector clients.

In the public sector, the company currently partners with Natural Resources of Canada, the US Geological Survey, the Alaska Geological Survey and the Israeli Geological Survey.

In these partnerships, SeismicAI integrates its software into either existing early warning systems operated by these agencies or, in the case of Alaska, operates as the sole software based on existing seismic networks operated by these agencies.

On the private sector front, SeismicAI is making significant inroads in Central America, Europe and Asia. In Mexico, they’re in the process of deploying their solution to 250 branches of the country’s largest bank.

SeismicAI has raised more than $7M from VCs and family offices to date, and according to Sasson, the company will become profitable in 2025.

Founded 10 years ago, SeismicAI is now focusing on incorporating more advanced AI models to further improve accuracy and provide even earlier warnings.

“We are starting to put much more emphasis on AI models, implementing AI in ways that will increase the accuracy of the system and will allow us to offer more lead time for our customers,” Sasson says.

“We’re talking about seconds: if you can give customers four, five, six more seconds of warning — that can be dramatic.”