Drownings are surging in Australia. AI can help

This year, Australian drowning deaths reached their worst level in three decades. Tragically, 357 drownings were reported between July 1 2024 and June 30 2025, with many more non-fatal incidents.

Australian drowning fatalities have surged because of “crisis-level” declines in swimming skills, especially among regional, remote and migrant communities.

Swimming at unpatrolled beaches and inland waterways that typically don’t have lifesaving services has also contributed to these deadly trends. So too has rock fishing.

With people now off work and enjoying the summer holidays, drowning risk is even higher than normal.

Education and awareness remain tried and tested ways to reduce drowning fatalities.

For example, this summer, Surf Life Saving Australia’s virtual “Beach Passport Campaign” is enabling the public to easily locate patrolled beaches. But water safety experts are also partnering with computer scientists to harness the power of algorithms and artificial intelligence (AI) to help save lives.

High-tech watchtowers

Surveillance AI for lifesaving is one key advancement.

Cameras at coastal hazard sites (selected based on historical incident data) capture continuous video feeds. These are then analysed by AI to identify emergency events.

The advantage of Al drowning detection is in reducing emergency response times in dynamic environments.

For example, if a rock fisher was washed off the rocks and into the water, AI identifies the event and alerts emergency services (within seconds) so they can validate the emergency and deploy rescue resources.

Smartphones and citizen scientists

Rip currents are strong, narrow, fast-flowing currents of water that occur on many beaches and are difficult for swimmers to identify.

Thousands of images containing rip currents are needed to train an AI-based model.

Australian researchers are leading this endeavour with a huge repository of images collected through CoastSnap – a community-driven initiative where beach goers can take an image of the beach with their smartphones.

In partnership with Surf Life Saving Australia, RipEye is helping train lifeguards in rip detection and lifeguards are helping train AI.

A smartphone app available to the general public is also being developed for the future. At a beach location users can capture real time wave movements and currents with the app signalling if swimming conditions are safe.

AI aiding pool safety

Public swimming pools are important community assets providing health, social and economic benefits.

With 421 million visits annually, and councils increasing access through campaigns such as $2 entry, lifeguards are on high alert.

To aid pool safety, local councils are investing in cameras, sensors and AI algorithms to monitor pools, identifying potential drowning incidents and alerting lifeguards in real time.

Using overhead cameras to continuously monitor swimmer activity, AI algorithms are trained through machine learning to detect signs of swimmer distress. These can include prolonged submersion or erratic movements.

Lifeguards scan wide, complex scenes while managing glare, noise, heat, rain, crowds and fatigue. Smartwatch alerts fed by AI can significantly enhance detection.

Like other operators in fields such as emergency services, defence and aviation, lifeguards already receive training in scanning techniques, hazard recognition and decision-making under pressure.

But even highly trained operators are subject to the fundamental cognitive and perceptual limits of the human brain.

Read more: Why should my child take swimming lessons? And what do they need to know?

Designing information for the human brain

To work effectively, an AI-powered alert system must consider several basic human-centred questions.

For example, what information should be shown? Too much detail overwhelms; too little is ignored.

How should the information be presented? Visual cues (such as text, shapes, colours, icons, motion, flicker) or audio tones, or subtle vibrations each have advantages and drawbacks.

Where should alerts appear? On a smartwatch, a wall display, augmented-reality glasses? Poor placement can block vision or pull attention away from other incidents.

When should alerts appear? Too early and people may attend to other things; too late and the chance to intervene is lost.

These design decisions matter because drowning detection is a vigilance task – a type of attention known to decline rapidly under fatigue or stress.

Not a perfect solution

The advantage of AI drowning detection is in reducing response times in dynamic pool and beach environments.

However, this requires nearby and available rescue resources.

Many of the technologies involve camera-based image analysis, which can also introduce privacy concerns.

Another issue is that AI drowning detection is imperfect. It sometimes produces false alarms and fails to detect people in distress.

To overcome this, it’s important that people are not just properly trained in how to use AI to detect drownings, but also that an AI system clearly explains how an assessment was made, and adapts when errors occur so that lifesavers work with the system as partners and not simply as responders.

The benefit of AI alerts depends on human operators’ ability to interpret, trust and act. So the question isn’t just whether AI can detect danger – it’s whether the information it delivers is cognitively digestible when lifesavers need it.

But AI-powered lifesaving methods are no substitute for swimming skills. So if you do cool off with a swim this summer, always remember: never swim alone, swim between the flags, listen to lifesavers and be vigilant around the water.