🔍 What’s in today’s flow
💧 Orange transforms fiber optic cables into distributed acoustic sensors that detect water leaks within one meter of the cable using AI algorithms.
🤖 Automated Machine Learning (AutoML) achieves 98.2% accuracy in forecasting harmful algal blooms, outperforming traditional deep learning with just 600 seconds of computing.
📊 Anthropic's Claude chatbot experiences "unprecedented demand" and global outage amid Pentagon controversy, highlighting AI infrastructure fragility.
🔧 Flowless Octopo, a conversational AI agent for water utilities, delivers 9% water loss reduction by translating network data into natural language insights.
⚡ AI-generated deepfakes drove over $200 million in election-related financial scams in Q1 2026, with experts warning of escalation in upcoming elections.
AI RESEARCH SPOTLIGHT: AutoML framework achieves breakthrough in algal bloom forecasting
Researchers have applied the Fast and Lightweight Automated Machine Learning (FLAML) framework to forecast harmful algal blooms in drinking water reservoirs - a first for this technology in water quality management.
Key findings:
Achieved R² of 0.982 for blue-green algae and 0.991 for chlorophyll-a prediction
Outperformed deep learning models (LSTM, CNN-LSTM, Transformer) with only 600 seconds of computing
Tree-based ensemble learners capture threshold-driven bloom dynamics effectively
Reduces expertise barrier for utilities implementing predictive systems
Why it matters
This approach democratizes advanced water quality forecasting, enabling utilities without specialized machine learning staff to deploy early warning systems that protect drinking water supplies from harmful algal blooms.
CASE STUDY: Orange Fiber Detection: turning telecom cables into water leak sensors
What happened
Orange, in partnership with VIAVI and Lightsonic, has developed "Fiber Detection" - a system that transforms existing fiber optic networks into distributed acoustic sensors for water leak detection. Showcased at Mobile World Congress 2026, the technology uses light backscattering and AI algorithms to detect vibrations caused by leaks. Lightsonic's specialized AI analyzes acoustic signatures to pinpoint leak locations within one meter along the cable and within ten meters of its path.
Why it matters
This approach eliminates the need for battery-powered IoT sensors every 100 meters, leveraging existing telecom infrastructure instead. For water utilities, it offers a cost-effective monitoring solution across vast networks. The technology also has potential applications for detecting gas leaks, seismic activity, and ground movement - creating new revenue streams from existing fiber assets.
👉 Full article
LATEST IN AI: Claude chatbot surges to number one amid Pentagon controversy and global outage
What’s new
Anthropic's Claude AI experienced record-breaking demand in early March 2026 following a public dispute with the Pentagon. CEO Dario Amodei's refusal to permit Claude's use for mass surveillance or autonomous weapons led to a federal blacklisting. Paradoxically, the controversy propelled Claude to the top of Apple's app store charts, surpassing ChatGPT. Anthropic reported daily sign-ups quadrupling since January and free active users increasing by over 60%. The surge culminated in a global outage on March 2, 2026.
Why it matters for the water sector
This incident reveals the fragility of AI infrastructure under demand spikes - a critical consideration for water utilities integrating AI into operations. Utilities relying on cloud-based AI services should evaluate redundancy strategies, consider on-premise alternatives for critical functions, and ensure service level agreements account for unexpected outages. The controversy also highlights emerging tensions between AI ethics and government applications that may affect future water sector AI deployments.
AI TOOL OF THE WEEK: Flowless Octopo
Overview
Flowless Octopo is a conversational AI agent built specifically for water utilities. Sitting atop Flowless's WaterCloud platform, Octopo allows operators to ask natural language questions about network data and receive instant, data-backed answers. The tool transforms complex hydraulic data into actionable insights, helping field teams prioritize leak surveys, generate reports, and calculate ROI on demand. A pilot with SABESP Brazil achieved 9% water loss reduction.
Key features
Natural language queries for network diagnostics
Automated report generation and ROI calculations
Water-specific AI trained on utility operations data
Category scores (1-5)
Ease of use: 5/5 – Natural language interface requires no technical training
Cost: 3/5 – Enterprise pricing; ROI dependent on utility size and NRW rates
Security and privacy: 4/5 – Cloud-based with encryption; utilities should verify data sovereignty requirements
Integration: 4/5 – Hardware-agnostic; connects to existing SCADA and meter infrastructure
Overall (16/20)
Octopo represents the shift from dashboard visualization to insight generation in water management software. Ideal for utilities seeking to reduce non-revenue water without hiring specialized data scientists.
THE SHADOW OF AI: AI deepfakes threaten 2026 UK elections with financial scams and data poisoning
Issue
A report from the Alan Turing Institute warns that AI-generated deepfakes have evolved beyond mere misinformation, now enabling financial scams tied to elections. In Q1 2026 alone, deepfake-driven fraud caused over $200 million in losses globally. Attackers have also deployed "data-poisoning" attacks on chatbots, planting false stories to distort AI-generated election information - with tests showing 17% of chatbot answers amplified fabricated narratives.
Why it matters
While not directly targeting water infrastructure, these techniques could easily adapt to undermine public trust in utility communications or smart meter data. Deepfake scams impersonating utility officials could target customers during rate changes or emergencies. Water utilities should implement verification protocols for official communications and monitor for AI-generated impersonation attempts.
Takeaway message
AI's potential for deception extends beyond politics - critical infrastructure communications require new authentication standards.
Thanks for reading! I hope you’ve enjoyed this week’s edition and look forward to seeing you next week!
Dr. Andrea G.T