The AI in the Room Is Speaking Lashon Hara — and You Asked It To
Let's start with something that happens all too often.
Last week a manager asked an AI to help write a reference for a colleague. Midway through the prompt, the manager typed something like: "He's great, but between us, he can be a bit disorganized — don't mention that."
Of course, AI obliged; it wrote a glowing letter without mentioning the disorganization. For one brief, shameful moment I thought: wow, this thing is so helpful.
Then our parasha hit the manager over the head.
📜 The holiness code didn't come with a carve-out for software
Parashat Kedoshim opens with the most audacious command in the Torah: Kedoshim tihiyu — "You shall be holy, for I, the Lord your God, am holy." (Vayikra 19:2). Not: you shall be holy when it's convenient. Not: you shall be holy except in group chats and DMs.
The Ramban, famously, argues that kedoshim tihiyu is a meta-commandment — a directive to elevate yourself even in the spaces the Torah doesn't explicitly regulate.¹ The Torah gives you specific laws, but holiness is the spirit that governs the gaps. The Ramban calls the person who ignores this a naval birshut haTorah — a scoundrel with the Torah's technical permission.
That is, I submit, a fairly accurate description of some of what we do with AI.
Nothing in the Shulchan Aruch explicitly prohibits prompting a language model to help you gossip more efficiently. And yet...
🗣️ Lo telech rachil — but what if you outsource the walking?
Sitting right in the middle of the holiness code, sandwiched between "don't steal" and "love your neighbor," is the prohibition that anchors the entire laws of speech: Lo telech rachil b'amecha — do not go about as a talebearer among your people. (Vayikra 19:16)²
The Chafetz Chaim, in his exhaustive codification of these laws, defines lashon hara with clarifying strictness: even true information. Even well-intentioned sharing. Even things said with a smile.³ The test is not your mood; it is the effect on the subject. Does this damage their reputation, their relationships, their livelihood? Then it is lashon hara — full stop.
Now here is the question the Chafetz Chaim did not live to answer, but which I think about constantly:
If I ask an AI to help me craft a message that subtly undermines a competitor, am I the one walking as a rachil — or is the machine?
The answer, I think, is obvious, and slightly uncomfortable. Shlucho shel adam kemoto — a person's agent is like the person himself (Kiddushin 42b).⁴ The machine is my shaliach. I am the sender. I am responsible for what my agent delivers.
The AI does not have a yetzer hara. It has no inclination to gossip, to one-up, to wound. That is entirely my contribution to the collaboration.
🤖 But here is where it gets genuinely interesting
The flip side is real and worth sitting with.
AI, used with intention, is perhaps the most powerful shmirat halashon tool ever invented — because it does not get swept up in the moment the way humans do.
When I am angry and drafting an email I will regret, my AI will not match my energy. It will not add a zinger. It will not say "you know what, you're right, he IS insufferable." It will flatten my affect, smooth my rough edges, and occasionally ask me — with infuriating calm — whether I want to "soften the tone."
That pause. That tiny, maddening suggestion. That is tzelem Elohim reflected back at me through a machine that doesn't even know what tzelem Elohim means.
The Talmud teaches that the seal of the Holy One is emet — truth (Shabbat 55a).⁵ A well-prompted AI, trained to be accurate and balanced, will often drag us toward emet when our first instinct is something considerably messier.
The AI will not add a zinger. That pause — that tiny, maddening "soften the tone?" — that is kedoshim tihiyu as a push notification.
✨ When the tool becomes a tikkun
Here is what I keep coming back to, and what I think is genuinely underappreciated in most conversations about AI and ethics: this technology, wielded with the right intentions, is not merely neutral. It can be actively redemptive.
Think about what becomes possible when AI is pointed at chesed rather than cheshbon.
WHAT RESPONSIBLE AI LOOKS LIKE IN PRACTICE
A retiree in Cleveland uses AI to translate her late grandmother's Yiddish letters — letters that sat in a shoebox for forty years — into English, so her grandchildren can finally read them. Lost voices, recovered. Memory, preserved.
A community rabbi uses AI to prepare accessible Torah summaries for congregants with learning disabilities, giving every person in his shul — regardless of background — a seat at the table of Jewish learning.
A researcher uses AI to surface connections between a 12th-century Rambam commentary and a modern medical ethics question, in twenty minutes instead of twenty weeks. The halacha moves faster. Lives may be saved.
A grieving family uses AI to transcribe and organize hundreds of voice messages from a father who passed away, turning scattered audio files into a coherent record — a kind of digital sefer zikaron — that his children will carry for generations.
None of these are science fiction. All of them are happening now, in 2026, by people in communities like ours.
The Talmud in Sanhedrin teaches that one who saves a single soul is as if they saved an entire world.⁶ We usually apply this to dramatic moments — a rescue, a intervention, a life-saving act. But I wonder if it also applies to the quieter salvations: the voice recovered, the text made accessible, the connection made possible across time and language and disability.
AI did not conjure those outcomes. The person who chose to use it that way did. The tool is inert. The intention is everything.
This is, I think, exactly what kedoshim tihiyu is asking of us in our moment. Not to fear the new technology, and not to embrace it uncritically — but to bring our full moral selves to it. To ask, before every prompt: am I walking as a rachil here, or am I walking toward something holy?
The gap between those two paths is not the AI's to navigate. It is ours.
🧂 The practical theology
So where does this leave us? Somewhere genuinely useful, I think.
Kedoshim tihiyu is not just a prohibition. It is an aspiration architecture — a framework for building a life that trends upward even in the unmarked territories. AI is, right now, a largely unmarked territory. The halachic apparatus hasn't fully caught up. The communal norms are still forming.
Which means we get to build them. And the raw material is already in the parasha.
Before you send a prompt, ask: am I walking as a rachil here? Am I outsourcing something I'd be embarrassed to do myself? Is my AI agent about to do something that, if I announced it at a Torah reading, would cause me to redden?
When AI pushes back, or softens, or gently asks whether you want to reconsider — consider that maybe the machine is, in this one moment, modeling something holier than what you typed.
The Torah does not ask us to be perfect. It asks us to be kedoshim. Holy — which is to say, striving, self-aware, and occasionally embarrassed by our first drafts.
In that sense, AI might be the first technology in history that, used well, can serve as a mussar shmuess with infinite patience and no memory of last week's failures.
That is not nothing. In fact, used with the right intentions, it might be everything.
🍽 SHABBAT TABLE QUESTION
When was the last there was something meaningful — a family memory, an act of chesed, a piece of Torah — that you've been putting off, that AI could help you finally bring into the world?
Shabbat Shalom 🕯️
— Fred
Fred Goldman is Managing Partner at Gently Ventures, (including bavl.pro + Jerrypress.com), father to eight awesome kids, and a serious Sefarim geek.
📎 Endnotes
¹ Ramban on Vayikra 19:2 — the concept of naval birshut haTorah. The Ramban argues that one can technically observe every explicit commandment while still behaving in a way antithetical to holiness, hence the need for this overarching meta-directive. Sefaria: Vayikra 19:2
² The prohibition of rechilut and lashon hara derives from Vayikra 19:16, Lo telech rachil b'amecha. Sefaria: Vayikra 19:16
³ Chafetz Chaim, Sefer Chafetz Chaim, Introduction and Hilchot Lashon Hara 1:1 — lashon hara applies even to true statements when they cause damage. Sefaria: Chofetz Chaim
⁴ Shlucho shel adam kemoto — Talmud Bavli, Kiddushin 42b–43a. The principle of agency and transferred responsibility. Sefaria: Kiddushin 42b
⁵ "The seal of the Holy One is emet" — Talmud Bavli, Shabbat 55a. Sefaria: Shabbat 55a
⁶ "One who saves a single soul is as if they saved an entire world" — Talmud Bavli, Sanhedrin 37a. Sefaria: Sanhedrin 37a
