Putting Israel on Trial with AI Witnesses
The machinery is new; the appetite for evidence against Jews is ancient.
For a prejudice that claims to uncover what others are afraid to say, antisemitism has always required an extraordinary quantity of invented evidence. The medieval blood libel accused Jews of murdering Christian children for religious purposes, furnishing persecution with the agreeable conviction that it was protecting the innocent. Generative artificial intelligence offers this habit of mind an improved production facility. The witness need never have existed. The photograph need never have been taken. The accusation can arrive complete with the evidence commissioned to support it.
To understand what this threatens in Israel, consider the moral preparation required before the murder of a civilian can be greeted with satisfaction. The victim must first be deprived of individuality, absorbed into a category whose supposed guilt makes further inquiry unnecessary.
An Israeli eating breakfast, attending a concert or putting a child to bed must somehow become an agent of oppression whose death can be accommodated within the language of liberation. Here the propagandist performs his indispensable service for the terrorist. He supplies the explanation that will survive the discovery of the bodies. Synthetic images and invented testimony could make that service cheaper, faster and more persuasive. But the decisive transaction remains human: the spectator accepts permission to regard another person’s slaughter as the settlement of an account.
When an accusation flatters its audience’s conception of its own goodness, what incentive remains to examine it? And when doubt threatens expulsion from the company of the virtuous, how many will insist upon doubting?
The blood libel’s enduring resource is this capacity to recruit moral feeling into the service of persecution. Its contemporary adaptation need not mention religion. It requires only that cruelty be treated as an expression of Jewish character, that Israeli citizenship constitute a permanent indictment, and that Jewish objections be admitted solely as further evidence of manipulation. Such reasoning possesses a remarkable immunity to correction. Every accusation confirms the verdict; every rebuttal enlarges the conspiracy. What possible evidence could acquit the accused?
Precisely because the historical comparison is grave, it requires discipline. Arab Palestinian hardship cannot be dismissed, and allegations against Israeli forces must be examined. Criticism becomes something else when responsibility migrates from particular decisions and identifiable actors to Jews collectively. Defending Israel means resisting that migration without granting its government immunity from scrutiny.
Artificial intelligence adds another device to the accuser’s equipment: the ability to invoke fabrication whenever Jewish suffering becomes inconvenient. The Anti-Defamation League (ADL) documented unsupported claims that photographs of the October 7 aftermath were AI-generated. The possibility that images can be forged becomes sufficient reason to reject unwelcome evidence. Thus a technology capable of inventing witnesses can also be invoked to silence actual ones. Israeli victims acquire the additional obligation of proving their suffering to people whose political satisfaction depends upon disbelieving it.
An effective Israeli response must make evidence available for independent examination: original recordings, corroborated testimony, accessible archives and corrections when claims fail. Platforms should expose coordinated fabrication; editors should investigate provenance before rewarding outrage with publication. Supporters of Israel who circulate convenient falsehoods damage the standards on which their arguments depend. A democracy defending its citizens has every reason to insist upon those standards, including when they produce uncomfortable findings.
No political grievance confers permission to murder Israeli civilians, and no fabricated image can make that permission respectable. Those who exploit humanitarian language to excuse terrorism should be required to explain where, exactly, their concern for human beings expires. The technological question concerns what machines can manufacture. The civilisational question concerns what people are willing to believe, and whom they are willing to abandon, for the pleasure of feeling righteous.
Whatever artificial intelligence learns to counterfeit, the moral choice will remain with the person who looks at an Israeli child and decides whether that child’s humanity requires a political qualification. Behind every life that propaganda reduces to an accusation lie the ordinary intimacies hatred must teach us to overlook: a familiar footstep, an unfinished conversation, someone expected home. The blood libel has always depended upon persuading people that their cruelty honours a higher duty. We should recognise its achievement whenever the murder of a Jew becomes an occasion for eloquence about everything except the murdered person. A machine can manufacture the image; only a human being can welcome the permission it appears to give. And no sophistication of language, no magnitude of grievance, can absolve us of what we choose to do with that permission.
