Inside AquAI - Your weekly briefing on AI in water #32

💧  Researchers achieve 99.28% accuracy with TinyML water quality monitoring system costing under $80, enabling decentralized safety monitoring

🤖 K-water showcased award-winning AI-powered water management technology at Mobile World Congress 2026 in Barcelona.

📊 OpenAI released GPT-5.4 with Codex Security, offering new capabilities for analyzing complex systems and code vulnerabilities.

🔧 Autodesk Info360 Asset now integrates AI-driven CCTV analysis to streamline pipe inspections and rehabilitation planning.

⚡ The International AI Safety Report 2026 warns that AI-generated deepfakes disproportionately target women and girls worldwide.

Researchers from LORDS Institute of Engineering and Technology and affiliated institutions in India developed a smart water quality monitoring system that integrates low-cost sensors with TinyML (machine learning running directly on a microcontroller). Built around an ESP32 device, the system continuously measures pH, turbidity, temperature, and total dissolved solids (TDS) and analyses the data locally using a neural network to detect contamination events.

Unlike traditional monitoring systems that rely on cloud processing, the AI model runs directly on the device, enabling real-time anomaly detection even in areas with limited connectivity. Data can also be transmitted to a cloud dashboard for remote monitoring and visualization.

This research highlights the potential for low-cost edge-AI monitoring networks that could expand water quality surveillance, particularly for rural water systems, decentralized treatment plants, and communities without access to continuous laboratory monitoring.

👉 Full paper 

Korea Water Resources Corporation (K-water) showcased three breakthrough technologies at Mobile World Congress 2026 in Barcelona. Their Smart Water Network Management (SWNM) system won a Global Mobile Award, marking a first for a Korean public institution. The technology uses IoT sensors to reduce leaks and stabilize pipeline pressure across distribution networks. K-water also demonstrated AI-powered autonomous water treatment plants and digital twins that replicate dam operations in virtual environments.

This deployment proves AI water management has reached commercial maturity. K-water signed partnership agreements with European organizations including the Catalan Water Partnership, opening pathways for global technology transfer. Water utilities worldwide can now access proven solutions for leak reduction, treatment optimization, and real-time decision support.

👉 Full story  

OpenAI released GPT-5.4 on March 5, 2026, featuring enhanced reasoning capabilities and a new Codex Security module in research preview. The model includes improvements to multi-step problem-solving, with the accompanying GPT-5.4 Thinking System Card documenting how reasoning models now better control their chains of thought. Codex Security specifically targets code vulnerability detection and security analysis.

Water utilities increasingly rely on SCADA systems and operational technology that require robust cybersecurity. Codex Security could help utilities audit control system code, identify vulnerabilities in legacy infrastructure software, and assess risks before deploying new digital solutions. Enhanced reasoning capabilities also support complex hydraulic modeling and scenario planning for drought or flood conditions.

👉 Full story 

Autodesk Info360 Asset integrates AI-powered video analysis with traditional asset management for water infrastructure. The platform uses vapar.solutions Technology to automatically review CCTV pipe inspection footage, identifying defects, cracks, and deterioration patterns. What previously took months of manual review can now be completed in hours, allowing utilities to create rehabilitation plans faster and prioritize critical repairs.

Ease of use: 4/5 - Intuitive interface designed for water professionals; minimal training required for basic functions.

Cost: 3/5 - Enterprise pricing model; costs offset by significant time savings on inspection analysis.

Security and privacy: 4/5 - Cloud-based with SOC 2 compliance; data encryption and role-based access controls.

Integration: 4/5 - Seamless connection with Autodesk's water portfolio, plus open APIs for CMMS and enterprise system.

Overall (15/20): Info360 Asset addresses a critical bottleneck in infrastructure management, the time required to analyze inspection data. For utilities with aging pipe networks and limited staff, AI-assisted analysis can accelerate condition assessments and help justify capital improvement budgets with data-driven evidence.

🔌Try it

The International AI Safety Report 2026, authored by over 100 experts including Turing Award winner Yoshua Bengio and backed by 30+ countries, reveals that AI-generated deepfakes are becoming increasingly realistic and harder to detect. The report documents that these synthetic media disproportionately target women and girls for harassment and non-consensual intimate imagery.

For water utilities and public services, the report highlights broader systemic risks: AI-enabled cyberattacks on critical infrastructure are increasing, with criminal groups using AI to identify vulnerabilities and write exploit code. Labor market disruption could affect utility workforce planning. The report urges organizations to adopt "defense-in-depth" approaches layering multiple safeguards against misuse.

Water sector leaders should assess their AI governance frameworks now, before capabilities outpace protections.

👉Full story

Thanks for reading! I hope you’ve enjoyed this week’s edition and look forward to seeing you next week!