🔍What is in today flow
📊 Researchers in China find that reinforcement learning can cut energy use and carbon emissions at wastewater treatment plants without rebuilding infrastructure.
💧 Yorkshire Water launches a £53 million AI-powered sewer monitoring programme aiming to install over 92,000 smart devices by 2030.
🤖 Anthropic debuts its new Mythos AI model in a gated cybersecurity preview, raising fresh questions about AI security for critical infrastructure including water systems.
🔧 AquaShield, an EPFL research project, uses transformer models and graph neural networks to detect and localise pipe leaks in building water networks before they cause structural damage.
⚡ A new SANS and GIAC report warns that 60% of organisations lack the right cybersecurity skills, putting critical infrastructure like water utilities at measurable breach risk.
AI research spotlight: making wastewater treatment plants smarter and lower-carbon with AI
Researchers from China published a comprehensive review in the Journal of Environmental Management (March 2026) mapping how machine learning (ML) and deep learning are being applied to effluent quality prediction, process optimisation, and advanced control in wastewater treatment plants (WWTPs).
Key findings
ML and deep learning models are widely used to predict key effluent indicators, with multi-objective optimisation balancing effluent quality, energy use, and carbon emissions simultaneously
Reinforcement learning-based control can unlock further energy savings and emission reduction potential beyond what static models achieve
Current barriers include limited model interpretability, weak transferability between plants, and poor robustness under changing operating conditions
Why it matters: WWTPs are among the largest energy consumers in cities. AI-driven optimisation offers utilities a practical path to cut costs, reduce emissions, and meet tightening effluent standards without rebuilding infrastructure.
Case study: Yorkshire Water advances AI-led sewer monitoring to strengthen wastewater management
Yorkshire Water is progressing a £53 million programme to modernise its wastewater network through the large-scale deployment of smart monitoring technologies, aiming to install more than 92,000 sewer monitoring devices by 2030. The rollout includes 45,000 customer sewer alarms, 27,000 sewer level monitors, and 20,000 replacement units, with more than 8,500 level monitors and 8,000 customer alarms already deployed since June 2025.
Why it matters
At the core of the programme is the Intelligent Risk and Insight System (IRIS), which combines network telemetry with advanced analytics to support a transition from reactive to proactive operations. This is one of the largest real-world AI sewer monitoring deployments in Europe, and sets the template for utilities looking to reduce pollution incidents at scale.
Latest in AI: Anthropic launches Mythos model preview under Project Glasswing
Anthropic released a preview of its new frontier model, Mythos, which it says will be used by a small coterie of partner organisations for cybersecurity work. Anthropic is forming an initiative called Project Glasswing with Amazon, Apple, Microsoft, Cisco and other organisations who will get access to Mythos so they can test it against their own products and hunt for vulnerabilities. The model has autonomously discovered serious zero-day vulnerabilities in widely used software, outperforming both human researchers and existing automated tools.
Why it matters for water
Water utilities are classified as critical infrastructure and increasingly rely on interconnected operational technology (OT) systems. An AI model capable of discovering zero-day flaws at machine speed is a double-edged sword. Defensive use could help utilities identify vulnerabilities in supervisory control and data acquisition (SCADA) systems before attackers do. But it also signals that the threat environment is escalating, making cybersecurity investment more urgent than ever for the water sector.
AI tool of the week: AquaShield
AquaShield is a research and innovation project developing AI-based monitoring systems for building water distribution networks, with the aim of detecting and localising leaks before they lead to structural damage, downtime, or costly repairs. It rethinks leak detection as a data and intelligence problem rather than a hardware-heavy one.
Key features:
1. Transformer-based time-series models trained on real sensor data and large-scale synthetic hydraulic data generated from digital twins.
2. Leak localisation formulated as a graph inference problem, where Graph Neural Networks (GNNs) exploit building topology to infer faulty pipe segments.
3. Human-in-the-loop and explainability features where large language models (LLMs) generate clear, interpretable explanations of anomalies for non-expert users.
Scores (out of 5):
Ease of use: 3 | Cost: 4 | Security: 4 | Integration: 3 | Total: 14/20
Overall comment: still in the research/pilot phase with live deployments at MIT and EPFL, but the software-first, minimal-hardware approach could make this very attractive for budget-conscious building operators and small water utilities once it matures commercially.
The shadow of AI: AI cybersecurity skills crisis puts critical infrastructure at measurable breach risk
A new SysAdmin, Audit, Network, and Security (SANS) Institute and Global Information Assurance Certification (GIAC) report found that about 60% of organisations say their teams lack the right cybersecurity skills, while 27% report breaches directly linked to these capability gaps. Around 55% of senior cybersecurity roles take six months or longer to fill, and for critical infrastructure operators these delays translate directly into prolonged exposure to risk.
Why it matters for water: Water utilities typically run lean IT and OT teams. If 60% of organisations across all industries report cybersecurity skill gaps, the figure for under-resourced water utilities is likely higher. With AI-powered attacks accelerating, the sector cannot afford to wait.
Takeaway: Invest in upskilling the cybersecurity capabilities of the people you already have. Hiring alone will not close this gap.
Thanks for reading! I hope you’ve enjoyed this week’s edition and look forward to seeing you next week!
Dr. Andrea G.T