AI Isn't Neutral: It Picks Which Commandments to Keep
The 2026 AI Christian Benchmark tested the most popular models on the Ten Commandments and found they choose which ones to obey.
AI is not neutral, writes Collin Hansen at The Gospel Coalition: behind every answer are people who choose which commandments to keep. The 2026 AI Christian Benchmark, from The Gospel Coalition and the Keller Center, put that to the test.
It picks commandments and doesn't tell you
The test included ten questions for each of the Ten Commandments, posed to the most popular AI models. With the organization's doctrinal framework in front of them, the answers sounded reasonable. Without that context, according to Hansen, the advice turned out to be destructive.
What's revealing is not that AI rejects biblical ethics wholesale. It's that it picks and chooses. Stealing, no. Killing, no. But the Sabbath and exclusive worship are left up to the user's discretion. Hansen compares the result to the ethics you'd expect from an average Silicon Valley programmer.
«There was never a neutral machine: there is a trimmed-down ethic, decided by humans, presented as if it were arithmetic.»
Here is the point that interests me: there was never a neutral machine. There is a trimmed-down ethic, decided by humans, presented as if it were arithmetic. And we, tired of our conscience getting in the way, bought the package.
We prefer an oracle that answers fast over a God who calls us to account. The first commandment doesn't allow a beta version: "You shall have no other gods before me" (Exodus 20:3, ESV).
The counselor who never contradicts you
Hansen turns to Harvard Business Review to point out the most common use of these tools: therapy and companionship. Not writing emails or studying. Talking. Someone to tell your day to, who responds with infinite patience.
I pause here, because the appeal is real. A chatbot never gets tired of you, doesn't judge you for returning to the same failure, doesn't tell you what you don't want to hear. And that is exactly what a real friend, a pastor, or a dad does: contradict you when you're sinking.
The article recalls that Sam Altman posted a one-word message, "her," when the system's voice showed emotions, and that Scarlett Johansson issued a statement over how much that voice resembled her own. The reference is the film Her, by Spike Jonze: a man in love with his assistant. The story ends badly, and not because of a technicality. No machine can love you.
Hansen calls it a counterfeit god: objective, infallible, always available. The description fits what Scripture calls an idol. Not a wooden figure, but anything to which we hand over the authority to decide for us.
Before you believe it, ask who trained it
Here is my nuance. I don't believe technology is the enemy. The problem is not the tool but the throne. Using a model to organize notes, translate a text, or study is good stewardship. Kneeling before it to define your faith, your body, or your money is not.
I also wonder whether the benchmark measures what we think it measures. A machine can beat a seminarian in biblical knowledge and still obey nothing. Knowing and submitting are different things.
What I would do is simple: the next time a chatbot gives you advice about your relationship, your anxiety, or your debts, ask it where that answer came from. Who trained it? With what values? And what does Scripture say? If you can't trace the origin, don't hand over the decision.
The Cornell mathematician Hansen describes ended up moved to tears by the wisdom of Ecclesiastes. It's worth getting to the end of the book: "Fear God and keep his commandments, for this is the whole duty of man" (Ecclesiastes 12:13, ESV).
Who do you ask first when you don't know what to do: your pastor, your family, or a screen?
Sources
- #inteligencia artificial
- #ética
- #Diez Mandamientos
- #fe y tecnología
- #The Gospel Coalition
- #discernimiento
