With increased AI use will come increased cybersecurity risks to infrastructure, public health and safety. The possibilities are grim: Hackers accessing computers controlling treatment plants could shift the chemical mixture to poison people or shut off the water supply. More often, the news features hackers getting into electrical utility systems than water utilities, but the concern exists.
“Drinking water and wastewater systems are an attractive target for cyberattacks because they are a lifeline critical infrastructure sector but often lack the resources and technical capacity to adopt rigorous cybersecurity practices,” then-EPA Administrator Michael Regan and White House National Security Advisor Jake Sullivan wrote in a letter sent to all 50 governors in March 2024. The letter named recent and ongoing threats of cyberattacks to the drinking water system in the United States.
“There is always a risk,” says Brian Zavareh, the PhD student working on sewer line inspections. Small changes to how these models are programmed could create big problems: Picture retraining the program for an autonomous driving vehicle to view a stop sign as a yield sign. Industry standards are trying to head off concerns around data leaks and close cybersecurity loopholes — before AI solutions are added to a system.
Water users who have worked with new AI solutions encourage others to start with the basics. Before onboarding an AI vendor, ask how data is encrypted, stored and protected. Companies can secure and show proof of data certifications, such as ISO certification, a third-party validation that vets the security of data storage and encryption based on internationally recognized standards.
But AI users should begin taking caution when determining what information is released to an AI system. At Denver Water, a cybersecurity team scans the system for data that shouldn’t be there, traces it to its source, and re-educates the person who shared it about withholding confidential or restricted data, personally identifying information or other potentially compromising details.
Jonathan Spitze, director of business technology and project management with Denver Water, suggests focusing on using publicly available information or what would be subject to a Colorado Open Records Act request as the guideline for what to share with artificial intelligence. Because bad actors are already using AI to their advantage, increasing the number of cyberattacks, AI may also need to be recruited for defense, to detect and thwart cyber-attacks, suggests Ralph Erik Exton, executive director of the Water Environment Federation. The Department of Homeland Security’s Cybersecurity and Infrastructure Security Agency (CISA) has offered “cyber hygiene services” to help water utilities identify and reduce vulnerabilities. The American Water Works Association also offers a risk management tool for water sector cybersecurity. Request cybersecurity technical assistance or an evaluation here.
The risks associated with AI don’t always come from cyberattackers — humans must think critically about the inherent bias of AI and analyze any recommendations from an AI agent. AI’s biases emerge because human biases skew the original data used to train AI, according to IBM. This can perpetuate societal and cultural biases and generate inaccurate or undesirable results. For example, had an AI agent planned Denver Water’s Lead Pipe Replacement Program, AI may have prioritized pipe replacements based on parameters that did not take into account disadvantaged neighborhoods. How can inherent bias be avoided? Organizations must govern their AI by managing and monitoring AI activities.
Independent journalist Elizabeth Miller writes about environmental issues in the American West for publications including The Washington Post, Scientific American, Outside, Backpacker, and The Drake.
Print