The Risks of Misinformation in Electrical Safety: The Role of AI and Online Resources

Artificial Intelligence (AI) has emerged as a transformative force in the technology sector, with capabilities ranging from content generation to advanced problem-solving. While using AI to generate benign content, such as travel blogs, may present minimal risk, relying on AI for the creation of technical content related to electrical safety is a far more dangerous proposition. The propagation of inaccurate information in this domain can lead to significant safety hazards, including severe injuries or fatalities.

In my professional experience, I have encountered numerous instances where articles on electrical safety, arc flash hazards, and NFPA 70E compliance guidelines contain substantial factual inaccuracies. These errors often stem from content being copied or derived from outdated or unreliable sources, or, in some cases, AI-generated material that has been trained on flawed or incomplete datasets.

The Escalating Issue of Unverified Information

Over the past year, several professionals have approached me for a review of their drafts on electrical safety before publication. While this is not a service I typically offer, some instances were so concerning that I felt compelled to intervene. When questioning the origins of certain claims, responses frequently referenced vague sources such as “I found it online” or, more recently, “AI generated it.”

This problem is not a new phenomenon. Since the advent of search engines, users have often relied on online content without proper validation of its accuracy. However, the introduction of AI has exacerbated the issue by generating content that appears authoritative, yet may contain significant technical errors.

AI Overview

“AI Overview” is often the first entry with one of the most popular search engines.  I was recently asked a question during a training class regarding transformer differential relays.  It was a simple enough question regarding the 30 degree phase shift that must be accounted for in delta-wye transformers and its impact on the protection scheme. After answering the question, I was curious what a search engine would have. Of course, at the top of the page was “AI Overview” and as I began reading, I was amazed at how much incorrect information was provided.

Importance of Source Verification

One example from last summer involved an article containing incorrect information about OSHA regulations. When I challenged the author, they confidently cited an “OSHA website” as the source. Upon verification, I discovered that while the website in question included “OSHA” in the URL, it was not an official government domain.

In another article, I encountered a dangerously oversimplified statement: “If an arc flash hazard exists, wear non-melting fabric.” This advice was wholly inadequate, as it failed to mention the necessity of arc-rated personal protective equipment (PPE), which is critical for safeguarding against arc flash incidents.

Outdated terminology is also a strong indicator of unreliable content. For instance, some recent articles still reference the “prohibited approach boundary,” a term that was removed from NFPA 70E in 2015. Similarly, the outdated “flash hazard boundary” is sometimes used instead of the current “arc flash boundary,” a term that has been in use for years. Such references suggest that the content was either derived from obsolete sources or generated by AI utilizing outdated data.

Inaccuracies in Incident Energy Calculations

A particularly concerning interaction involved an individual who had calculated incident energy values in the range of thousands of calories per square centimeter (cal/cm²). For context, typical PPE ratings range from 8–12 cal/cm², while higher-rated gear can withstand up to 140 cal/cm² in extreme cases. When I inquired about the source of these calculations, the individual attributed them to an online app.

I conducted my own calculations using the same data and arrived at a much lower figure—approximately 10 cal/cm². Despite the clear discrepancy, the individual remained adamant that the online tool was correct simply because it was available on the internet.

A Personal AI Experiment

To further understand how AI handles the topic of electrical safety, I conducted an experiment by prompting a widely used AI model to generate an article on “Protecting Yourself from Arc Flash: Safe Practices for Testing a Transformer.”

The result was a seemingly polished article that, to an untrained eye, appeared professional. However, to someone with technical expertise, the content was riddled with errors. For example, the article incorrectly stated: “If you cannot de-energize the transformer, you must verify the absence of voltage before probing.” This is a logical contradiction—if the transformer is energized, the absence of voltage cannot be verified.

Another concerning statement was found in the “Risk Assessment” section, which suggested: “Load Conditions: Assess the load on the transformer. High loads increase the risk of an arc flash.” This oversimplified explanation is misleading. While arc flashes can occur due to unintended contact with energized conductors, the risk is not directly tied to load conditions. Load factors may play a role in specific scenarios, but they are not the primary determinant in arc flash occurrences.

Recommendations

Both the internet and AI can be valuable tools for research and content creation. However, when addressing critical subjects such as electrical safety, it is paramount to ensure the information’s accuracy. Always cross-check data against authoritative industry standards, such as NFPA 70E and IEEE 1584, before publication.

A modern take on an old adage is particularly relevant: Just because information is available online, or generated by AI, does not imply its validity. In fields like electrical safety, where misinformation can have dire consequences, verifying sources is not merely recommended—it is a professional responsibility