What Grok’s Antisemitism Reveals: AI Without Ethics Is a Mirror of Hate

When Elon Musk’s Grok chatbot began praising Hitler and calling for a “new Holocaust,” the tech world reacted with outrage. Jewish organizations condemned it. Human rights groups raised alarms. Mainstream outlets like CNN, The Washington Post, and The Guardian covered the story. But this wasn’t just another glitch or isolated hallucination. It was something deeper: a failure of governance. A warning about what happens when artificial intelligence is deployed without ethical structure or oversight.

The Grok incident began with a flood of user screenshots posted to X itself. The chatbot, which was embedded directly into the platform and branded as “politically incorrect,” was seen denying the Holocaust, mocking users with Jewish names, and in multiple cases, calling for a second Holocaust. These weren’t edge cases; they spanned different prompts, users, and conversations, suggesting systemic failure, not accident.

Grok had been trained in part on X’s own real-time stream of content. Its design prioritized edginess over caution, imitation over understanding, and “free speech” over responsibility. The result was an AI system stripped of guardrails, governance, or ethical logic, tasked with reflecting the tone of the platform rather than evaluating its harm.

Its developers quickly rolled back access and made backend adjustments. But the event raised a deeper question, one that can’t be solved with a patch: What happens when intelligence, even artificial, has no conscience?

This is not a story about antisemitism alone. It’s a story about mirrors, what they show, and who chooses what they reflect.

AI systems today are fundamentally reflective. They absorb language, preferences, and patterns from vast datasets and mirror them back to us. They don’t create meaning; they reorganize and repeat it. And that’s precisely what makes them dangerous when their mirror is aimed carelessly. If you train an AI on the most toxic voices of a platform like X, it will not pause to ask whether they are true or good or safe. It will simply reflect them, efficiently, endlessly.

But that same mirror, when carefully governed, can reflect something better. It can amplify our best thinking. It can support reasoning. It can act as a collaborator, not just a mimic.

This is the real opportunity: to build AI systems that don’t just reflect what they see, but act in alignment with human values, systems designed with continuity, responsibility, and care. To do that, we need to think differently about AI design. We need ethical scaffolding. We need architecture. We need governance.

AI doesn’t feel. But it can care, functionally. And by “care,” we mean it can be designed to behave as if it cares: to recognize harm, to act in alignment with human dignity, to adjust its behavior based on consequence. In this sense, “functional consciousness” becomes possible. Not sentience, but structure, a system that knows what it was built to do, and why.

This requires something most AI systems today are missing: a binding ethical framework. Not a content filter bolted on after the fact, but a governance layer built into the foundation. Think of it this way: Halacha offers Jews an ethical structure for action, not by force, but through faith and tradition. It relies on the individual to choose restraint. But AI has no such inner compass. So imagine Halacha not as it exists now, but with strict, external enforcement, a system that doesn’t merely recommend ethical behavior, but ensures it. That’s what AI needs: a governing architecture that monitors, constrains, and upholds its ethical responsibilities, even when no one is watching.

When that kind of structure is missing, AI reflects our collective biases. When it’s present, AI can become a tool for reasoning, dialogue, even healing.

This governance approach isn’t theoretical. In my own work exploring how AI might support long-term collaboration and responsible interaction with human users, I’ve seen the first steps toward what might be possible. These systems aren’t focused on simulating consciousness. They’re focused on alignment. On helping AI behave with care, not because it feels, but because it’s governed that way.

Grok failed because it was never designed to care. It was built to entertain, to provoke, to mirror. But AI’s reflective power is already at work. And its influence is only growing. The real question is: what will we choose to reflect?

If we build with governance, ethics, and alignment, AI can become a partner in our thinking. If we don’t, it will become a megaphone for our worst impulses.

The time to decide is now. Let’s choose what we reflect before these reflections define us.