AI’s expanding possibilities never cease to fascinate me.
Anyone who knows me knows how deeply involved I am with artificial intelligence. I use it every day, both in my professional career and in my Torah study and writing. I’ve written elsewhere about how AI is already transforming Torah learning, opening remarkable new paths for research and interpretation.
But for all the opportunities AI offers, I share the growing concern many people feel about where it might lead. Some worry about its effects on workers in the economy; others, about the risks it poses to society at large. And lately, with the release of powerful video-generation tools like OpenAI’s Sora and Google’s Veo, a new anxiety is rising: We may soon live in a world where it’s nearly impossible to tell what’s real and what’s fake. Authentic-looking images, voices, and videos can now be created from nothing, blurring the line between artificial and real—a prospect many find terrifying.
Andrew Bosworth, Meta’s chief technology officer, spoke about this issue a few months ago.
I talk often with people about this one, and it’s a hard one to wrap their heads around, but I have to remind people that actually the period that we grew up in was very unusual historically. Before the photograph, and before video, all media was presumed to be possibly fake, right? Like letters, the newspapers, were presumed—you know, you didn’t know the veracity of it. There was a very unique period, never probably happening again, where you could produce a piece of media—a photograph or a video—that it was impossible to imagine faking it. It was just orders of magnitude more expensive to fake than to have it be real. And so these were presumptively true. That’s not going to be the case anymore. So we’re going to return to a, like, pre-1900s media relationship that we have with media.
Bosworth didn’t go much further than that, but for me, his point carried a deeper truth: Our brief modern faith in photographs and video was a historical anomaly, a century-long pause in humanity’s natural skepticism. Before cameras, every document and image could be forged, and people knew it.
The Torah never left that world. It always lived with the awareness that seeing alone isn’t believing. What is seen must be reported, tested, and defended by accountable witnesses.
The Torah’s law of witnesses never trusted evidence or objects; it sought truth through human accountability, consequence, and cross-examination. As AI forces us to return to this pre-modern state of skepticism, this system offers more than just a historical parallel but a blueprint for rebuilding trust in how we determine what is true.
A Pre-Digital Skepticism
The Torah never assumes that evidence speaks for itself. It insists instead on the moral accountability of human testimony:
A single witness may not validate against a person any guilt or blame for any offense that may be committed; a case can be valid only on the testimony of two witnesses or more.
That verse is procedural, yet it reveals a deep philosophy: Truth doesn’t come from objects or recordings. It comes from people who are willing to stand behind their words.
The Talmud illustrates this principle with a striking example. Imagine you saw someone chase another into an abandoned building, then found the pursuer holding a blood-stained sword over the dying victim. It seems obvious what happened. Yet Sanhedrin 37b teaches: “If that is what you saw, you saw nothing.” Circumstantial evidence, no matter how strong, cannot convict; only two witnesses who directly saw the act and can be questioned are able to establish the event.
To modern readers, that rule long seemed incomprehensible. Contemporary courts can convict or hold someone liable with surveillance footage and audio recordings. Why rely on fallible human memory when a camera can show what happened? Yet halacha has never accepted such media as primary evidence. A beit din still requires living witnesses whose accounts can be questioned and compared.
What once appeared rigid or outdated now feels almost prophetic: In the age of AI-generated images and voices, the Torah’s insistence on human testimony feels like the surest path to truth.
The Fragility of Memory
To be fair, human testimony isn’t perfect either. Modern psychology confirms what the rabbis already understood—that memory is flexible and vulnerable. Cognitive scientist Elizabeth Loftus showed that every time we recall a memory, we partially rewrite it. In the Deese–Roediger–McDermott experiments, participants falsely “remembered” hearing words that were never spoken, because their minds supplied related ones that fit the pattern.
A 2023 Guardian article described how our brains generate “short-term memory illusions,” filling in missing visual details. Another study at the University of Sussex found that we reshape experiences into coherent narratives.
If memories can change, doesn’t that make witnesses unreliable? In a way, yes. But the halachic system never expected perfection—it anticipated human weakness. It disqualifies pesulei edut—biased or related witnesses—and penalizes false testimony. Its strength lies not in assuming people never lie, but in structuring scrutiny so that truth still emerges.
Two witnesses, questioned separately, are unlikely to distort an event in the same way. Derisha ve-chakira (cross-examination), based on Deuteronomy 19:18 tests perception and consistency. Truth, in this view, is not something one person observes—it’s something multiple people build together through examination.
When Cameras Lie
For most of modern history, we trusted machines more than people. A photograph couldn’t forget; a recording couldn’t lie. Judges treated such media as self-authenticating.
But that trust has evaporated. Legal scholar John Villasenor warned that deepfakes —AI-generated videos designed to mimic real footage— could make us doubt even genuine videos, eroding trust in what we see. The Illinois State Bar Association now cautions that even genuine recordings may be dismissed as false. The National Law Review recently wrote that in today’s climate, “no foolproof method currently exists to classify text, audio, video, or images as authentic or AI-generated,” raising the real possibility that genuine recordings may be doubted.
The phrase some ethicists now use is the “liar’s dividend”: the idea that once convincing fakes exist, liars can dismiss even genuine evidence as fabricated. Once evidence becomes infinitely malleable, the camera loses its authority.
What remains standing is the witness: the human being willing to stand by their word.
When Witnesses Replace Cameras
As images and recordings lose their authority, we return to human witnesses, not because people are flawless but because they can be questioned, compared, and held responsible. Jewish law has always known this and built safeguards into its system. It requires two witnesses, not one, to prevent collusion, and it obligates judges to question them closely to expose inconsistencies and test reliability. The system rests on accountability: Words carry weight, and false testimony is treated as a grave offense.
The Torah commands, “You shall not swear falsely by My name, profaning the name of your God,” making honest testimony a sacred duty. The oath is not only religious; it’s psychological. It ties words to conscience and consequence, reminding the witness that speech carries meaning. The threat of perjury is a serious deterrent; few people will risk the punishment or disgrace that comes with being caught in a lie. Yet the oath also draws on something deeper: a human desire for integrity, an aversion to living with a troubled conscience, and a shared understanding, whether grounded in divine commandment or moral instinct, that truth matters.
That ancient legal culture may offer the guidance we need now: a renewed ethic of verification that values human accountability over mechanical certainty.
Seeing Anew
The Torah’s narratives often show the limits of sight alone. Eve saw that the fruit was beautiful, but not the danger it held. Isaac’s blindness allowed Jacob to deceive him. Halacha turns this moral insight into law: Witnesses must report what they have seen, and judges must probe what they have heard.
Modern life reversed that instinct. We learned to trust what we could see, to take images and recordings as proof of what is real. But as AI reshapes what can be seen, we’re reminded that sight alone is never enough, and that truth depends on questioning what is seen and conveying it with integrity.
If the age of AI returns us to an older skepticism, the halachic world never left it. In that sense, the future of truth may look surprisingly ancient. As AI erodes the trust we once gave cameras and microphones, the halachic system—built on verification and cross-examination—offers a well-tested blueprint for rebuilding trust in an era where seeing is no longer believing.
When Andrew Bosworth said we’re “returning to a pre-1900s media relationship,” he likely meant it as a warning. But it may yet prove an opportunity.
