Inside aquAI - 20 Jan 2026 | Edition No. 26

Flow time: 5 min I your weekly pulse on AI news, tool and case studies reshaping the water sector

💧 New Aalto University research finds DeepSeek R1, ChatGPT-4o, and Gemini 2 all successfully answer complex water science questions, with DeepSeek R1 leading in three out of four technical domains tested

🏭 A US wastewater plant cut aeration energy use by 30% using AI-powered process control from Xylem

🔒 Signal co-founder launches privacy-focused ChatGPT alternative that doesn't train on user conversations

⚠️ ChatGPT accounts for over 60% of enterprise generative AI data risk incidents, raising security concerns for utilities

📚 Scroll.ai emerges as a knowledge management tool that turns company documents into searchable AI assistants

Aalto University researchers evaluated DeepSeek R1, ChatGPT-4o, and Gemini 2 across four water science domains: machine learning and optimization, remote sensing, flood modeling, and sediment transport. Using a novel scoring rubric assessing relevancy, accuracy, authenticity, and novelty, all models successfully addressed benchmark questions.

Water professionals can confidently use current AI models for technical support across specialties. Understanding each model's strengths, DeepSeek for novelty and most domains, ChatGPT for remote sensing, Gemini for speed, helps utilities choose the right tool for specific workflows and technical challenges.

👉 Full paper

Moxie Marlinspike, Signal's co-founder, has launched a privacy-focused alternative to ChatGPT that promises not to train AI models on user conversations. The platform uses end-to-end encryption principles and allows users to run queries without surrendering data ownership. Unlike mainstream chatbots, it prioritizes data protection over model improvement, giving users control over sensitive information.

Water utilities handle confidential operational data, infrastructure details, and customer information that shouldn't be shared with third-party AI platforms. A privacy-conscious chatbot could enable engineers and operators to use AI assistance for asset management, compliance reporting, and troubleshooting without exposing sensitive utility data to training databases or potential security breaches.

👉Full story

US wastewater treatment plant deployed Xylem's AI-powered aeration control system to optimize dissolved oxygen levels in real time. The technology uses machine learning algorithms to analyse influent characteristics, weather patterns, and biological activity, then automatically adjusts blower speeds and valve positions to maintain optimal treatment efficiency while minimizing energy consumption.

Aeration typically accounts for 50-60% of a wastewater plant's energy costs. This deployment demonstrates how AI can deliver immediate operational savings while maintaining compliance. For utilities facing rising energy prices and tighter budgets, predictive process control offers a clear path to reduce costs without compromising treatment quality or environmental performance.

👉Full story

Scroll.ai transforms company documents, SOPs, technical manuals, and institutional knowledge into an AI-powered search assistant. It creates a private knowledge base that employees can query using natural language, making it easier to find answers buried in PDFs, spreadsheets, and archived reports without uploading data to public AI platforms.

Overall 14/20 – - Strong fit for utilities with extensive technical documentation, standard operating procedures, or compliance manuals. Helps operators quickly find answers to operational questions without digging through filing systems, though initial document organization effort is required.

🔌Try it

ChatGPT accounts for over 60% of enterprise generative AI data risk incidents, according to new security research. Employees frequently paste sensitive information into the chatbot, including proprietary data, customer details, and confidential documents, creating significant data leakage risks that bypass traditional IT security controls.

Water utilities manage critical infrastructure information, customer data, and operational details that could pose security risks if exposed. When staff use public AI tools for help with reports, troubleshooting, or analysis, they may inadvertently share information about system vulnerabilities, treatment processes, or network configurations that should remain confidential.

Establish clear AI use policies and provide approved tools before employees create their own workarounds.

👉Full story

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