UL Solutions experts Bruce Johnson and Robert Marshall said on May 21 that the fire service must prioritize safety and human oversight when using artificial intelligence (AI). In a recent article, they outlined both the opportunities and risks of integrating AI into daily operations, especially as AI-generated information increasingly shapes training, planning, and compliance decisions.
The topic is important because fire service decisions rely on accurate, up-to-date information. Codes and best practices change frequently in response to new risks. Johnson and Marshall said that AI tools can sometimes provide outdated or incorrect guidance without indicating when their information is no longer current. “Many AI tools produce responses that sound clear and authoritative but are not fact-checked or current. They can reflect outdated codes, ignore recent changes or apply general guidance incorrectly. Because AI-generated information is presented with confidence, it can be difficult to recognize when it is incomplete or wrong,” they said.
According to the article, AI is already used informally across the fire service for tasks such as studying for promotional exams, drafting training programs and policies, summarizing code requirements, generating checklists for inspectors and marshals, and organizing research. Even without formal adoption by agencies or departments, these outputs are influencing real work tied to compliance efforts.
Emerging hazards like battery energy storage systems further highlight these challenges. The experts explained that real-world incidents have led to significant updates in safety guidance for such systems; however, “AI tools can potentially generate responses based on earlier or outdated assumptions about these systems.” This means there could be critical gaps if firefighters rely solely on unverified outputs from AI.
Broader industry efforts are underway to address reliability concerns related to AI products. Initiatives like UL 3115 focus on evaluating how human oversight should be integrated into the use of AI-based products so that risks from incorrect outputs can be managed effectively.
Johnson and Marshall concluded: “AI is already influencing how information is created and trusted. To support safety and maintain public trust, its use needs to be guided by clear expectations, validated against current standards and grounded in professional judgment.” More details about independent evaluation services for safer deployment of AI-enabled products are available through UL Solutions.
