AI makes information abundant but it makes human judgment scarce. Scarce things are usually worth protecting.
Here’s the key point: Jewish practice (study, fixed commitments, community) protects the habits that make human judgment possible, such as thinking through hard issues, sticking to unpopular stands, and doing all that with and for other people—what we might simply call human agency. Agency is exactly what AI is likely to erode.
AI is more likely to lead us astray through drift than through malice. With time, the machine gets better at producing reasonable answers, and we get lazier about doing the work that once made us reasonable people. Questions stop forming because answers arrive too fast. Effort seems wasteful because tools do the hard parts. Social interaction becomes optional because a simulation is cheaper than showing up. We can try to regulate potential dangers at the margins, but we can’t legislate the willingness to think, to follow through, to be serious, to be present. If those habits atrophy, nothing explodes; everything just thins. It looks as if it’s working—until it isn’t.
Judaism’s answer to this kind of debilitating drift is neither alarm nor nostalgia. It’s discipline: a set of practices that add friction where it forms character, and remove it where it hinders participation.
Start with Shabbat, the weekly refusal to optimize. One day a week, the signals we usually obey—pings, dashboards, convenience—are silenced. This is not a productivity hack; it’s an obligation. Obligations outlast convenience. By keeping them, we practice two scarce skills at once. First, we practice choosing where to put our attention. There’s no infinite scroll at the table; you pick what to notice and hold it there. Second, we practice being there. You eat, sing, talk, and walk without device props.
Judaism is a mimetic tradition—passed down through lived practice, not just texts. Replace human mentors with AI, and you flatten tradition into homogenized certainty.
Then come study and argument. Serious Torah learning is not just content intake. It’s the slow, stubborn habit of supplying and testing reasons. You sit with a sugya that won’t yield quickly; you argue it out with someone who will not let you cheat; you revise your theory and try again. You join a discussion that has been ongoing for a few thousand years and to which you need to pay careful attention to follow the argument. AI can fetch sources and summarize moves, but it cannot give you the reflex that keeps moral talk from devolving into sentiment.
If automation hollows out jobs, what will people do all day that feels meaningful? A recognizable Jewish answer has been around for a long time: strenuous study in the beit midrash as a vocation, bounded by daily and weekly commitments and tethered to other people—teachers above, students below, peers alongside. The work is hard and cumulative; the day is bounded by prayer and the week by Shabbat. In a world where many jobs turn into “manage the agent that does your job,” Judaism’s answer isn’t escapism, but sanity. The structure and the substance give you meaning, even if, or maybe precisely because, you’re not optimizing the usual things.
Nonnegotiables—kashrut, fixed prayers, life-cycle duties—are scheduled repetitions of doing what you said you’d do even when you don’t feel like it. That’s how commitments beat preferences. Structured interdependence—minyan, shiva houses, and schools that function only if neighbors show up—means coordination is not a heroic improvisation but the default setting. And a class of goods sits outside the market by design: Torah and mitzvot are not for sale. That’s a boundary, not sentiment. The boundary protects the thing from the optimization pressure that dissolves it.
Can all this survive AI? More precisely, can we use AI to widen the on-ramp to Torah without hollowing out the work these practices protect?
AI can help, at least at first. AI knocks down barriers that kept too many Jews out of serious learning until now. You can surface sources in seconds, translate on the fly, follow cross-references that once took a small lifetime, connect with a chavruta across oceans, and help overworked teachers prepare a better shiur in less time. People with jobs and kids and little background now have a real on-ramp. Good. Use it.
But the same affordances that open the door can thin the workout. The slope is gentle enough to look harmless. First the tool finds sources—helpful. Then it summarizes—fine. Then it analyzes—tempting. Then it recommends conclusions—convenient. And then the part that changes you quietly stops happening. The beit midrash looks busy, while the habit of agency quietly evaporates.
We don’t need bans. We need friction in the right places. Make it easy to get to texts and people; make it impossible to leave with unearned certainty. Map sources, line up arguments, pose the right questions—excellent. Stamp answers? No—not beyond those that are well-established and noncontroversial and serve as the hard foundation for engagement and discussion. Make checking easier than trusting. Show real disagreement instead of compressing it into consensus. Build for two.
This can sound like piety until you translate it into product choices. Giving AI users a tour of the relevant sources rather than a neat conclusion hints at what matters. A “think before you peek” delay is a way to reintroduce the pause that makes judgment possible. The metric for AI builders shouldn’t be time to answer; it should be time spent in conversation with texts and with a partner.
To be sure, sports, the army, orchestras, serious craft—each builds attention, follow-through, and coordination. The difference is coverage, integration, and depth of commitment. Judaism runs the training weekly, daily, seasonally, and across a lifetime and ties it to texts and other people. No talent gate, no admissions office, just unending commitment. You show up and submit to the rhythm, and you get the benefit of doing the work. That scaffolding is exactly what consumer technology erodes and Judaism protects.
Will any of this add up to a religious revival? I am not in the prophecy business. I will say that technologies that drain something human may create demand for practices that restore our humanity. If AI makes passive competence cheap and omnipresent, human agency becomes scarce. And people, being people, start looking for it. There are already tentative signs of change. Families without prior ritual life discover that a screens-down Friday night gives them one block of time the world cannot colonize. Two beginners in different cities schedule a weekly online discussion with a text open before them and find that the commitment, and maybe the text, changes them. None of this proves anything, but it shows a direction. When the problem is drift, the countermove is constraint. Judaism has the right constraints available.
Of course, there are easy ways to get this wrong; one is of immediate concern, while another is a bit more remote for now.
The first is the faux-rabbi product, a pat answer dispenser. It grows quickly because closure sells. Communities that adopt it are liable to become overconfident and incurious. Within a generation, you train people who can cite but can’t reason. There’s another cost. Judaism is a mimetic tradition—passed down through lived practice, not just texts. Rabbis don’t just answer questions; they transmit subtle judgment, show how to adapt principles to circumstances, embody living tradition. Replace human mentors with AI, and you flatten tradition into homogenized certainty. I know these dangers well because my nonprofit, Dicta, built just such a rabbi-in-a-box; we’ve learned from the experience, and the second version will focus more on supplying and organizing primary sources and less on drawing conclusions.
The second trap is the clever Shabbat hack—some technique that gets around Shabbat prohibitions to allow access to “indispensable” AI tools—preserving the surface while smuggling weekday pressures back in. You gain convenience but lose the only day of the week that trains you to resist convenience.
Both temptations, the oracle that replaces engagement and the hack that defeats Shabbat, are understandable. But both sabotage the human capacities we’re trying to defend.
What to do instead follows from the design principles noted above. Builders should measure success by how much thinking users can’t skip. If customers ask for an oracle, point them back to sources and questions; if they want a “Shabbat mode” that defeats Shabbat, say no. Rabbis and teachers should use the software to widen the on-ramp and then make the room harder, not easier—more text, more argument, less performance. Model how to hold real disagreements without anxiety and how to leave a question open without calling it a bug. Communities should adopt tools that respect their constraints and decline the ones that flatter them with instant certainty.
The work is the point. Take the help of AI. Just keep the part that cannot be automated: human agency. When information gets cheap, human agency becomes the luxury good. Judaism makes that luxury good available to everyone. Our job is to keep it available while the machines make everything else free.
Professor Moshe Koppel
