A.I. and Electrical Safety: A potentially dangerous combination

By Jim Phillips | Jan 15, 2024
Having A.I. create an article about favorite places to visit is quite different than creating a piece about electrical safety. Incorrect information could lead to injury or worse.

Artificial Intelligence (A.I.) has been dominating the tech news, with plenty of buzz about automatically generating articles, books and more. However, having A.I. create an article about favorite places to visit is quite different than creating a piece about electrical safety. Incorrect information could lead to injury or worse.

I have read articles about electrical safety, arc flash and NPFA 70E that contain many incorrect statements. It appears they are either cut-and-pasted inaccurate or obsolete information from the internet, or written by an A.I. app that likely pulled from a data set that included inaccurate information.

On several occasions just this past year, I was contacted by various people asking if I could provide a quick review of their drafts about electrical safety before they were published. I don’t normally do this. However, on several occasions the drafts were so alarming that I felt compelled to step in. When I asked about the source of some of the information, the cringeworthy response was “the internet.” And now, an A.I. app. This isn’t a new problem, as people have been using information found online since the early days of search engines, but now A.I. has taken it to a whole new level.

Please verify your sources

One article I reviewed over the summer was making incorrect OSHA statements. When I questioned the person, the answer was, “I got this from an OSHA website.” Surprised, I checked out the website in question, and although “OSHA” was in the URL, it was far from being an official administration website.

Another article I read that had already been published stated that if an arc flash hazard exists, wear nonmelting fabric. That’s it. No mention of arc-rated PPE.

Although not as potentially hazardous, use of outdated terminology can offer a clue about the accuracy. A few 2023 articles included references to the “prohibited approach boundary,” which was removed in the 2015 edition of NFPA 70E. The term “flash hazard boundary” can still appear even though this was changed to “arc flash boundary” long ago. You get the idea—the list goes on.

Incident energy calculations

I had a conversation with someone who calculated the incident energy in the thousands of calories per square centimeter (cal/cm2). For perspective of that eye-popping number, basic PPE is rated 8–12 cal/cm2 with higher arc ratings upward of 65, 75 and even 140 cal/cm2 for extreme cases. Of course, I asked how this value was determined. The response was, “I found an app on the internet.” I ran my own calculations with their data and found the incident energy was closer to 10 cal/cm2—a significant difference. The person argued that it was from an online app, so it must be correct.

My own A.I. experiment

I conducted my own experiment and asked a popular A.I. program to write a short piece about electrical safety. To someone with a limited knowledge about the subject, the result seemed very authoritative. However, if you are knowledgeable about the subject, it was anything but.

The topic for the article was: “Protecting yourself from arc flash: Safe practices for testing a transformer.” I admit the title is a bit misguided, as you should not test a transformer energized. One of my first forensic cases in the 1990s was exactly that, and the person conducting the testing did not survive.

Although the app-generated article was very well written, there were many inaccuracies.

Verifying absence of voltage is a great item to include, which is part of establishing an electrically safe work condition. Except the actual language was: “If you cannot de-energize the transformer, you must verify the absence of voltage before probing.” I’m sure that if you cannot de-energize, then there won’t be an absence of voltage.

There was also reference to “risk assessment,” another important topic. Under this heading, the A.I. article states: “Load Conditions: Assess the load on the transformer. High loads increase the risk of an arc flash.” That’s a new one for me, since an arc flash is typically caused by inadvertent contact between energized conductors, although a small number might also be attributed to equipment failure or other unusual causes.


The internet and A.I. are valuable resources to assist in writing an article. However, when it comes to subjects such as electrical safety, it is important to verify the information is correct based on the latest editions of electrical safety standards such as NFPA 70E and IEEE 1584.

A very worn-out phrase still applies, and I modified it slightly: Just because it is on the internet or generated by an A.I. program does not mean it is correct. / VectorMine

About The Author

PHILLIPS, P.E., is founder of and provides training globally.  He is Vice-Chair of IEEE 1584 Arc Flash Working Group, International Chair of IEC TC78 Live Working Standards and Technical Committee Member of NFPA 70E.  He can be reached at [email protected].






featured Video


New from Lutron: Lumaris tape light

Want an easier way to do tunable white tape light?


Related Articles