SAIL (Surf AI for Lifesaving) is Surf Life Saving NSW’s applied artificial intelligence program focused on one outcome: reducing the time between a person getting into trouble in the water and a lifesaver or lifeguard intervening.

Drowning is a time critical problem. In dynamic coastal environments, human surveillance is constrained by distance, glare, crowd density, fatigue, and competing operational demands. SAIL addresses this gap by augmenting, not replacing, lifesavers with persistent, automated detection capability.

The program uses shore based camera systems combined with computer vision models trained on real surf conditions. These models continuously analyse the water to identify high risk behaviours and patterns such as swimmers in rip currents, prolonged submersion, and abnormal movement trajectories. When credible risk is detected, alerts are delivered to operational staff, enabling faster verification and response.
SAIL is not a standalone technology experiment. It is integrated into Surf Life Saving NSW’s operational ecosystem, including communications, incident management, and rescue workflows. Design priorities include reliability in harsh coastal conditions, explainable alerts, controlled false positive rates, and usability under pressure. Performance is assessed against real rescue outcomes, not theoretical benchmarks.
Critically, SAIL shifts surveillance from reactive to proactive. Instead of relying solely on a raised arm or a bystander call, the system identifies risk before distress becomes visible or irreversible. The time gained, often seconds, is decisive in water rescues.

The program is already delivering measurable impact, with multiple rock fishing rescues initiated or accelerated by AI detections. These are operational interventions, not simulations.

SAIL demonstrates a pragmatic model for AI in public safety: narrowly scoped, outcome driven, ethically deployed, and embedded within human decision making to improve public safety outcomes.

Background:
Rip currents are Australia's leading coastal hazard and are responsible for more drowning deaths annually than floods, bushfires, cyclones and shark attacks combined. Despite sustained education campaigns, many beachgoers – particularly young males, culturally and linguistically diverse (CALD) communities, and visitors to unpatrolled beaches – continue to struggle to identify rip currents. In support of the Australian Water Safety Strategy 2030, this project aims to improve rip current recognition through the development of RipEye, an AI-powered smartphone tool and online game designed to enhance public awareness and rip identification skills.

Methods:
An annotated dataset of approximately 10,000 rip current images, spanning aerial, fixed-camera and beach-level perspectives, has been assembled from CoastSnap citizen science observations and coastal imaging archives across Australia to train and validate an AI detection model for diverse Australian beach conditions. Focus groups with surf lifesavers and high-risk community groups are concurrently informing co-design of the tool and an interactive online game to maximise usability and educational value.

Results:
Testing of the rip current detection model demonstrates strong performance on individual smartphone image frames, achieving approximately 85% accuracy for aerial imagery and 81% accuracy for beach-level imagery. These results substantially exceed the current ability of beachgoers, with previous studies showing only around one-third can correctly identify rip currents. Preliminary testing also indicates further performance improvements when analysing sequential image frames (video), which will be incorporated in future development. The online educational game, Challenge the RipEye, has also been developed, enabling users to improve rip current literacy by comparing their performance against the AI model.

Conclusions:
RipEye represents a new approach to drowning prevention by combining AI, smartphone technology and gamification to strengthen rip current literacy. Following successful pilot demonstration, RipEye has the potential for national implementation through Surf Life Saving Australia and broader international application.

Title: From Data to Decision: How DECKEE is Powering Government Waterway Intelligence Across Australia

Background
Boating fatalities remain a persistent challenge across Australian waterways. The Australian Water Safety Strategy 2030 identifies "Aligning policies and partnership for change" as a national imperative, calling for whole-of-government approaches, scaled resources, and unified education efforts, particularly for boating and watercraft. Despite strong intent, waterway agencies have historically operated with limited visibility into recreational boating behaviour, making evidence-based policy and targeted intervention difficult. DECKEE was built to close that gap.

Description

DECKEE is a maritime safety and decision intelligence platform operating across Australia and the United States. The DECKEE Operations Platform (DOP) provides government waterway agencies with operational analytics, geospatial activity intelligence, and community communication tools. Current agency partners in Australia include Safe Transport Victoria, The South Australian Government, Western Australian Government, Marine and Safety Tasmania, the Gold Coast Waterways Authority, and the Department of Fire and Emergency Services (WA). The platform enables agencies to understand where and how waterways are being used, identify high-risk zones and behaviours, deliver targeted safety campaigns, and measure outcomes, replacing reactive incident response with proactive, intelligence-led safety management.

Lessons Learned

Agencies with access to behavioural data make faster, evidence-based decisions. Shared intelligence infrastructure reduces duplication across jurisdictions and creates conditions for national alignment. Community adoption of the consumer app is the engine that powers government data quality, demonstrating that public engagement and agency outcomes are mutually reinforcing, not separate workstreams.

Conclusions

Technology-enabled partnerships between government agencies and safety platforms represent a scalable model for advancing the AWSS 2030 imperatives. DECKEE's work across Australian jurisdictions demonstrates that decision intelligence infrastructure can unify fragmented policy efforts, target education more effectively, and ultimately reduce preventable drowning and injury in recreational boating.

Background: Coastal drowning is a critical public health issue in Victoria, Australia, particularly at identified drowning ‘blackspots’ such as the Mornington Peninsula and Bass Coast. Existing coastal safety signage plays an important role in risk communication; however, its effectiveness is often limited by low engagement, poor recall, and inconsistent comprehension among beachgoers, including multicultural communities. Emerging technologies such as virtual reality (VR) provide opportunities to test risk communication strategies in realistic but controlled environments before implementation. This study aimed to identify the most effective coastal safety signage design for high-risk beaches using community-informed co-design and VR-based testing.

Methods: A mixed‑methods approach involved community and stakeholder focus groups and VR testing to underpin signage co‑design. Six focus groups, including two multicultural community cohorts, explored how different user groups interpret warning signs, prioritise information, and make safety decisions at beaches. Three unique warning sign designs were subsequently evaluated across multiple VR beach environments. Participants assessed perceived danger, willingness to swim, confidence in decision-making, likelihood of seeking further safety information, and intention to use QR-code enabled safety resources in multiple languages. Quantitative analyses were conducted, supplemented by thematic analysis of open-ended feedback.

Results: Focus groups (n=51) identified the need for simplified messaging, stronger visual cues, clearer hazard descriptions, and emotionally salient content to support safer decision-making. VR testing (n=327) identified significant differences between the three signage concepts. Two signs demonstrated greater effectiveness compared to the third design, demonstrating lower willingness to swim and perceived safety in hazardous environments, greater confidence in safety-related decisions, and increased likelihood of seeking additional safety information.

Conclusion: Combining community co-design with VR provided an iterative evidence-informed approach to developing coastal safety signage. VR facilitated rapid, low-risk testing prior to real-world implementation. This mixed-methods model has strong potential for adaptation across other coastal and inland waterway settings.

Background

Recreational water activities such as snorkelling and diving are popular across Australia, yet critical safety information about coastal environments—weather conditions, hazards, entry points, and local safety guidance—is often fragmented across multiple sources. This information gap can increase risk for participants unfamiliar with local conditions, including tourists, new ocean users, and occasional participants.

To address this issue, the Australian Water Safety Strategy 2030 identifies the need to localise water safety efforts through grassroots initiatives and improved access to practical safety information.

Description

DiveScout is a digital platform created by divers for divers that aggregates environmental conditions, hazard awareness, site-specific guidance, and community knowledge to provide accessible safety information for snorkellers and scuba divers.

In 2026, DiveScout conducted a beta pilot in Victoria. The program engaged snorkellers and scuba divers to test the platform and provide feedback on usability, information needs, and how the tool influences trip planning and safety awareness.

Lessons learned

Early findings indicate strong demand for consolidated, location-specific safety information. Participants also highlighted the value of community-generated local knowledge, which is often difficult to access but critical for safe participation.

These findings suggest digital platforms could strengthen individual risk assessment capacity and support collaboration between recreational communities and water safety stakeholders.

Conclusions

Digital platforms such as DiveScout may represent a scalable tool for drowning prevention by improving situational awareness and supporting safer decision-making in coastal environments. At scale, this approach could also generate valuable data on where, when, and how people snorkel and dive in Australia; helping address information gaps and supporting more targeted water safety interventions.

Background: Drowning is a leading cause of unintentional injury in coastal environments, particularly at high-risk locations beyond flagged patrol areas. Social media and generative AI are emerging influences on recreational decision‑making and destination choice. Visually compelling and peer-endorsed content may alter perceived risk, and redirect visitation to hazardous or unpatrolled locations. This study examined how these information sources influence beach choice and risk perception.

Methods: A mixed‑methods approach included semi‑structured interviews with beachgoers (n=29), along with social media and generative AI content analysis. Interviews explored perceived susceptibility and behavioural intentions relating to beach visitation and safety. Social media posts were analysed to identify platforms, content characteristics, engagement metrics and safety or risk cues. Generative AI outputs were systematically reviewed for beach recommendations, safety messaging and representation of hazardous locations.

Results: While only 7(24%) interviewees reported using social media to select beaches, younger participants and first-time visitors were more strongly influenced by online content. Social media analysis of Mornington Peninsula beaches identified 66 posts generating 3,591 unique comments and 1,641 shares. Analysed content contained limited safety messaging; only 30% of Instagram, 6% of TikTok and no Facebook posts included safety or explicit risk information.
Generative AI analysis identified consistent promotion of high‑risk ocean beaches (e.g., Gunnamatta and Sorrento Ocean Beach) across models, with variable and often limited safety information. Some models provided general warnings; however, safety messaging was inconsistent, with several responses omitting hazards entirely or presenting high‑risk locations alongside safer beaches without clear distinction.

Conclusions: Social media and generative AI have a strong ability to influence beach visitation and shape perceptions of risk among beachgoers. Peer‑endorsed and algorithmically-generated content may reduce perceived susceptibility and severity, potentially weakening protective behaviours. Drowning prevention agencies should actively monitor and engage with emerging digital information systems to strengthen safety messaging and influence safer decision-making.