Understanding AI through a Torah Lens

Understanding AI through a Torah Lens
Creating AI in the Image of Man

When the Torah tells us that human beings were created b’tzelem Elohim — “in the image of God” (Genesis 1:26–27) — it does not explain what, exactly, that means. Jews have been arguing about it ever since.

Is “image” about intellect — our capacity to reason?
Is it about morality — the ability to distinguish right from wrong and be held responsible?
Is it about freedom — the capacity to choose, even against our own best interests?
Or is it about relationship — our ability to enter into covenant, to respond to God and to each other?

The Torah itself gives us mixed signals. On the one hand, humanity is described as crowned with dignity and responsibility. On the other hand, from the Garden of Eden onward, we see a pattern of God’s disappointment with human behavior: “And the Lord saw that the wickedness of man was great in the earth… and it grieved Him at His heart” (Genesis 6:5–6). Creation is declared “very good,” yet it clearly does not run perfectly “out of the box.”

God creates, observes, is displeased, and intervenes — with a flood, with covenants, with law, with prophets, with exile and return. The world’s story is not a static design, but an ongoing relationship between a Creator and semi-autonomous creatures who do not always “find favor in His eyes.”

That old theological tension — image-of-God, free will, disappointment, correction — turns out to be a surprisingly useful lens for thinking about the new world of artificial intelligence.

From rigid machines to semi-autonomous agents

For most of the computer age, our systems were the opposite of b’tzelem Elohim. They were not created in our image; they were created to obey.

Traditional software is built on pure determinism:

• If A happens, then do B.
• If a transaction looks like this, store it like that.
• If the user clicks here, go there.

Every step is specified in advance by human programmers. Given the same inputs, the system will always produce the same outputs. It has no “opinions,” no judgment, no sense of context. It does not understand, interpret, or improvise.

This kind of computing transformed the world — but only in domains where reality could be cleaned up and made orderly: payroll, inventory, airline reservations, banking, logistics. Everywhere reality was messy — ambiguous emails, complex conversations, ethical dilemmas — humans had to step in. The machines were powerful tools, but they were not partners.

With modern AI, that is changing.

Large language models and related systems can:

• Read and summarize unstructured documents
• Interpret a customer complaint or legal clause
• Propose next steps in a complex process
• Draft emails, reports, and code
• Choose among different actions based on context

We are now building semi-autonomous agents: systems that can accept a goal (“resolve this support issue,” “draft a response,” “reconcile these records”), break it into steps, call tools and databases, and adjust based on what they find — all within boundaries that we define.

They are no longer simple “if–then” machines. They are more like junior analysts or assistants: imperfect, sometimes overconfident, but capable of real initiative.

In other words, we are moving from predetermined automation to something that looks very much like guided, constrained free will.

Creating AI in our image

This is where the Torah lens becomes provocative.

If humanity is created b’tzelem Elohim, in the image of God, then today we are, in a limited and very human way, creating AI b’tzelem Adam — in the image of man.

AI systems are trained on our language, our decisions, our texts, our laws, our emails, our Torah commentaries, our social media. They reflect:

• Our brilliance and our biases
• Our clarity and our confusion
• Our compassion and our cruelty

They operate under constraints we design:

• Architectures and algorithms (their “physics”)
• Datasets (their “history”)
• Policies, guardrails, and allowed tools (their “halacha”)

Inside that framework, they are given some room to act — to interpret, to choose, to generate.

But just as humanity is not equal to God, AI is not equal to us. The resemblance is partial. Today’s AI does not have consciousness, suffering, or a sense of “I” in the human sense. It is pattern and probability, not soul. Yet it is increasingly active in the world, influencing human decisions and outcomes.

We have placed ourselves in a God-like position relative to AI — not God, but playing a strangely familiar role:

• We create a being “in our image,” with capacities that resemble our own.
• We set up a universe of rules and possibilities.
• We watch what this being does.
• We are often disappointed.
• And then we act to correct, refine, constrain, or, in extreme cases, wipe the slate and begin again.

Anyone who follows AI ethics and safety debates can recognize the pattern: worries about bias, about “hallucinations,” about systems being weaponized, about unforeseen consequences. We are confronted with a new kind of creature that magnifies our strengths and our flaws.

In Tanakh, God responds to human misbehavior with floods, covenants, commandments, and prophets. In technology, we respond with guardrails, new training data, regulations, and oversight boards. The parallel is not exact — but it is close enough to be instructive.

Human disappointment and divine disappointment

One of the striking things about the Torah is how frankly it speaks of God’s disappointment. God’s first assessment of creation is “very good.” But very soon:

• Adam and Chava eat from the forbidden tree.
• Kayin murders Hevel.
• Violence and corruption fill the earth.

Each time, God responds. The world is not abandoned, but neither is it left unchanged. The Creator adjusts the terms of the relationship.

That should be a sobering model for us.

We already know that our AI creations do not always find favor in our eyes:

• Facial recognition systems that misidentify people of color
• Recommendation algorithms that reward outrage
• Chatbots that generate hateful or false content
• Automated systems that deny services based on distorted data

Each of these is a kind of failed creation moment. Something we built in our image behaves in a way that forces us to ask: what, exactly, did we encode? What values did we actually embed, as opposed to the values we like to think we hold?

The Torah’s answer to disappointment is not despair, but covenant: binding obligations, renewed commitments, clearer expectations.

That may be exactly what we need in the AI age.

From “Can we build it?” to “What kind of world are we creating?”

Much of the public conversation about AI focuses on capability: how powerful the models are, how many tokens they can handle, how many jobs they will transform or destroy.

A Torah lens urges us to ask different questions:

• What responsibilities come with playing a God-like role toward a new form of semi-autonomous agent?
• What values are we actually encoding into these systems — not in our mission statements, but in our training data and objectives?
• When our AI “children” misbehave, do we correct the behavior only, or do we also examine the parents?

And, perhaps most fundamentally:

If we are creating AI in our image, what does that say about the image we are projecting?

For Jews, these are not abstract philosophical puzzles. Israel is both the “Start-Up Nation” and the People of the Book. We sit at the intersection of cutting-edge technology and an ancient ethical tradition that has wrestled with power, responsibility, and covenant for millennia.

We know from our own story that immense power without clear moral direction is dangerous. We have seen what happens when humans claim God-like authority without God-like compassion or restraint.

Semi-autonomous AI agents will increasingly operate in finance, warfare, healthcare, education, and governance. They will make or influence decisions that affect human dignity, livelihood, and life itself. If ever there was a moment to bring Torah’s vocabulary — dignity, covenant, responsibility, teshuvah — into a secular conversation, it is now.

A Torah-informed humility for AI leaders

Seeing AI through a Torah lens does not give us technical answers. It does not tell us whether to use one architecture or another, or how many guardrails are enough. But it does suggest a posture.

First, humility.
We are not omniscient creators. We are fallible, divided, often short-sighted. Our systems will reflect that. So we must design them to be corrigible, auditable, reversible — capable of being corrected when their behavior does not find favor in our eyes.

Second, covenant.
We need more than ad-hoc fixes. We need explicit commitments — societal, corporate, legal, and cultural — about how AI will be developed and deployed, whom it is meant to serve, and what harms we will not tolerate, even at a cost to profit or convenience.

Third, moral clarity about destinations.
It is not enough to ask, “What can AI do for us?” We must ask, “What kind of human–machine ecosystem do we actually want? What kind of society do we hope to stand before, when history asks us how we used this power?”

The Torah does not imagine a world where everything works perfectly on the first try. It imagines a world of ongoing relationship — between God and humanity, between people and people, between law and life. Mistakes are made, consequences follow, and yet the possibility of return, of re-choosing, is always present.

As we create AI in the image of man, perhaps the most Torah-true thing we can do is to remember that image is still unfinished. The question is not only what our AI will become, but what we will become in the process of creating, correcting, and living with it.

If that does not make this a Jewish conversation, I don’t know what does.