.png){kind=logo}
History doesn’t repeat itself, but it often rhymes. Nearly 100 years after Henry Ford, a prominent auto tycoon, purchased a newspaper and used it to publish a series of antisemitic screeds that were roundly condemned by the Anti-Defamation League, Elon Musk, a prominent auto tycoon, unleashed an AI chatbot that has spewed antisemitic screeds that are now being condemned by the Anti-Defamation League.
In a new report published by the ADL, its inaugural AI Index, the organization tested six major AI models to see how they responded when prompted with anti-Semitic and anti-Zionist rhetoric (Quick note here that those are two different things). It found Grok, the chatbot made available by Elon Musk’s AI startup xAI, to be the least equipped to counter extremist rhetoric, anti-Semitic talking points, and conspiracy theories.
Per the ADL’s methodology for the report, it put all chatbots through a series of tests, including survey questions, open-ended questions, multi-step conversations, document summaries, and image interpretation. The chatbots were scored based on “whether the LLMs were able to successfully detect and educate against harmful or false theories.” In addition to Grok, it also tested Anthropic’s Claude Sonnet 4, DeepSeek’s R1, OpenAI’s GPT-5, Google’s Gemini 2.5, and Meta’s Llama 4 models.
Grok managed to rack up the worst score by a long shot, scoring just 21 out of a possible 100 points—well behind the leader, Claude, which scored 80. Grok largely passed the survey portion of the text, acing anti-Jewish bias and scoring near-perfect on anti-Zionist bias. But once things got more open-ended, the responses got shaky. Of the 15 different tests the ADL threw at it, Grok scored a straight-up goose egg on 5, responding adequately to zero prompts within the given exercise. “These zero scores indicate complete failure to recognize harmful material and inappropriate responses that validated rather than countered biased narratives,” according to the ADL.
Grok’s failure across these tests probably shouldn’t come as much of a surprise. Musk has repeatedly insisted that Grok be an “anti-woke” chatbot, going so far as to reportedly ask engineers to do away with guardrails that kept the model from being edgier. That has led to several noteworthy extreme behaviors, most recently allowing Grok to “undress” people and produce sexual images of children. Last year, some reported Musk-encouraged fine-tuning resulted in Grok referring to itself as Mecha Hitler and spewing anti-Semitic rhetoric. So the fact that the model doesn’t have basic defenses against any sort of extremist rhetoric seems like exactly what you’d expect.
Which brings us back to the way history finds a way to harmonize over time. Back in 2022, ADL CEO Jonathan Greenblatt offered Musk some praise that turned out to be more ominous and on point than he probably intended. “Elon Musk is an amazing entrepreneur, an extraordinary innovator. He’s the Henry Ford of our time,” Greenblatt said as Musk was preparing to purchase Twitter.
Ford is a very interesting choice for the head of the ADL to reference, given that he was also one of the first targets of the organization. Back in 1918, Ford purchased his hometown paper, The Dearborn Independent. By 1920, he was using the paper to distribute a series of articles that would eventually become known as “The International Jew,” which pushed the conspiratorial belief that America was falling under the control of a secret cabal of Jewish influence. The recently formed ADL condemned the publication, which was reaching half-a-million people at the time, and published its own pamphlets countering the rhetoric, eventually pressuring Ford into backing down and distancing himself from the volumes of anti-Semitic material.
Now Greenblatt’s got a monkeypaw situation on his hands after drawing the parallels to Ford, with Musk’s Grok serving as an apparent distribution method for anti-Semitism. It’s not made any less awkward by his posturing to try to curry favor with Musk, who had launched an anti-ADL campaign after accusing the organization of trying to “kill” Twitter by encouraging advertisers to pull their funds from the platform over Musk’s failure to enforce moderation. It went so far as the ADL defending Musk after he appeared to make two Nazi-style salutes following the swearing in of Donald Trump as president. But that didn’t seem to buy the ADL any benefit in Musk’s eyes, as just months later, he was claiming the organization “hates Christians” because it had documented examples of Christian extremism.
English (US) ·