Can This App Predict Shark Attacks?

In 2023, South Australia experienced five shark attacks — four of them on surfers, with three proving fatal. For a region far less populated by surfers than Australia’s East Coast, the cluster was hard to fathom.

North Shore figure Tamayo Perry, 49, was surfing off Goat Island on the east side of Oahu last June when he was fatally attacked by a suspected tiger shark. His lifeless body was reportedly found by local surfers on the island, missing an arm and a leg. Shortly after, Australian freesurfer Kai Mckenzie, 23, lost his right leg to a three meter great white in Port Macquarie.

“With whites, none of the usual rules apply,” marine biologist Brinkley Davies told us. “They are stealth, attack predators and are often unpredictable. They’re like a bloody Volkswagen. They attack by stealth, so seeing one prior to an attack is incredibly rare. In many areas where great white sharks are found, such as South Australia, the underwater environment often consists of dark, weed-covered granite and rock. This makes it difficult to see below the surface, even in clear water. They’re only really visible from the side.”

Ultimately, a shark attack is a tragic event anyone would prefer to avoid.

As such, Evan Valenti, a lifelong Shark Week fan, AI hobbyist, and post-COVID surfer turned “solopreneur,” has developed a tool that aims to mitigate future attacks. Combining his passions, he has created SafeWaters.AI — a fledgling app designed to forecast the risk of shark encounters.

Valenti, who began surfing in Narragansett, Rhode Island, was inspired to develop the app after learning from a news headline that, “eight great whites had been tagged where I’d been trying to surf just days before.” Reflecting on that moment, Valenti recalls, “What if I could apply these AI algorithms to forecast what I dubbed at the time as ‘shark crime?’” The idea was built around a previous exercise where he’d used AI to predict the risk of crime in certain areas.

SafeWaters.AI uses machine learning to predict high-risk days for shark encounters by analyzing over 200 years of global shark attack and marine weather data — which amounts to about 0.0000444% of the 450 million years that sharks have been cruising the oceans.

Valenti explains, “Our model includes latitude, longitude, date, and roughly thirty marine weather variables for every day — whether or not an attack occurred — at every beach that has had an incident.”

The app’s AI model claims to have achieved an ambiguous 83% accuracy rate in assessing risk levels, helping beachgoers make informed decisions without causing unnecessary alarm. Valenti emphasizes that the app doesn’t predict shark attacks but instead forecasts the likelihood of one occurring in a given location.

“It’s going to work best at beaches that have had attacks in the past. If there’s never been an attack there, it’s probably going to read ‘low risk’ for most or all of the days,” Valenti explained to Newsmax last year. “That’s where the 83% (accuracy) stems from.”

He noted that some beaches with no attacks on record will occasionally show a medium or high risk day when nearby beaches, which have a history of attacks, show similar forecasts. “Extra precaution definitely can’t hurt, as every beach has its first incident eventually, especially with attacks on the rise and sharks coming closer to shore,” he told Stab. 

SafeWaters.AI said they aim to support marine conservation by committing to donate 5% of its profits to ocean cleanup initiatives.

While the app is free to download, a 7-day risk forecast costs $2/month. Valenti had initially intended the app to be free, but soon realized that this model would likely run the business into the ground.

The company expects seed funding in the coming months and is working to promote the tech within the surfing community through partnerships with professional surfers, who Valenti is currently unable to disclose.