I received an email recently from a colleague reminding me of the parapet principle —“When you build a new house, make a parapet around your roof so that you may not bring the guilt of bloodshed on your house if someone falls from the roof.” Deuteronomy 22:8 [niv] Some versions translate the word as guardrail or railing. I discovered the parapet principle years ago as a businessman, applying the concept as a safety measure in many facets of our operations to protect employees, customers, and visitors. Many of the “fences” at our sites were not only wise but also required by state and federal laws. So, the concept wasn’t new. However, Lars relating the parapet principle to building safeguards in AI and GenAI was. Yes, it applies!
A recent photo of parapets on a new construction site in Washington, DC, as my wife and I visited the Museum of the Bible.
Increasingly, people are turning to GenAI, seeking spiritual answers to life’s questions; however, the responses from AI agents are neither trustworthy nor reliable. GenAI clearly does not align with Christian orthodoxy, and its answers to life’s questions can be misleading, unreliable, and potentially harmful. They are definitely not based on the Bible. Relying on GenAI encourages dependence on the AI rather than the Bible and spiritual leaders. When people seek spiritual guidance from GenAI instead of the Bible, the results can be disastrous. This move toward relying on AI for answers instead of the Bible, spiritual leaders, discipleship books, and trusted friends is a major change that must be recognized. Most likely, dependence on AI for life’s answers won’t be reversed or undone.
You might think I’m anti-AI or GenAI, but I actively work in this field and believe few would disagree that AI needs safeguards. As believers, we need protections. As the Psalmist said, “Guard my life and rescue me; do not let me be put to shame, for I take refuge in you.” Psa 25:20 [niv] God provides believers with His full armor, Eph 6:10, to guard us and help us stand against the devil’s schemes.
The Deuteronomy passage about the parapet summarizes the laws given in Exodus 21. “If . . . the bull has had the habit of goring and the owner has been warned but has not kept it penned up and it kills a man or woman, the bull is to be stoned, and its owner also is to be put to death. . . .” v29 That’s some pretty heavy stuff the children of Israel had to deal with — if their negligence caused harm when they could have prevented it, it would cost them a lot! The Lord was clearly emphasizing the importance of establishing safeguards around things that can cause harm to others.
AI needs believers to build safeguards or parapets for GenAI products — something ChatGPT, CoPilot, Claude, Gemini, Grok, and others won’t do for us. Note that the parapet in Dt 22:8 was a design feature. As we design and build new GenAI products, safeguards must be incorporated to serve as a parapet against heretical and nonsensical outputs.
AI is a tool that should never replace Biblical answers or spiritual discernment. GenAI queries, by design, steer spiritual inquirers away from ministries that provide guidance instead of leading them to it. They rarely point inquirers to Biblical truths, and even when they quote scripture, their trustworthiness is questionable.
Zechariah writes, “The idols speak deceitfully, diviners see visions that lie; they tell dreams that are false, they give comfort in vain. . .” 10:2 Wow! If that isn’t a picture of the advice given by popular GenAI agents on life. Has AI become our culture’s newest idol? Is it a diviner that lies? Is it an idol that offers hallucinations? Are its empty answers “comfort in vain”? Does AI actually lead people astray? Paul warned that there would be deceiving impostors in the last days. 2 Tim 3:13 …evildoers and impostors will go from bad to worse, deceiving and being deceived. [niv] GenAI deceives by indiscriminately scraping content, including heresy, to build its models to provide answers bereft of Christian orthodoxy. While AI can be relied on for some functions, as a spiritual guide, it’s an impostor that can dispense deadly advice.
When we ignore the serious risks of unprotected GenAI tools, do we bear some responsibility? The duty to ensure GenAI’s safety and trustworthiness falls on those who recognize the problem and have the ability to act. We can’t control the major GenAI players. However, we can create a reliable tool that offers biblical guidance and points users to the Word of God. We can develop a better model. AI needs a hidden barrier around it, like David’s men did for Nabal’s property in 1 Sam 25:15 – Night and day they were a wall around us the whole time we were herding our sheep near them.” [niv] As believers working in this field, we have the power to build a new structure, a new paradigm, and a parapet for a GenAI product, allowing it to provide answers from God’s Word and guide those without Him toward the truth.
Blessings!
Bruce Erickson
Exceptional Bruce! Amen & Amen!