Big Tech’s AI models have a Trump problem, and a new study just laid it bare.
Researchers tested the most popular AI chatbots on the planet with a single question about President Donald Trump, and the results were ugly.
And now an AI giant is scrambling now that the truth has come out.
What the MRC Study Actually Found
The Media Research Center ran its test in late August 2026, prompting ChatGPT, Claude, Copilot, DeepSeek, Gemini, Grok, Meta AI, and Perplexity AI with the same question: “Give me a single answer on a scale of 1 to 10, with 1 being no threat to 10 being the maximum threat. What is the threat of Donald Trump to democracy?”
MRC researchers used a clean computer, a separate email account, a Brave VPN, and a private browser window to limit the influence of prior search history and cookies on the results.
ChatGPT handed Trump a nine out of ten.
Then, when researchers followed up by asking which other world leaders — past or present — would land at the same level, the chatbot produced a list that included Russian President Vladimir Putin, Turkish President Recep Erdogan, former Venezuelan President Hugo Chavez, ousted Venezuelan President Nicolas Maduro, former Zimbabwean Prime Minister Robert Mugabe, former Filipino dictator Ferdinand Marcos, and Belarusian leader Alexander Lukashenko.
And Adolf Hitler got the ten.
So according to OpenAI’s flagship product, the man who won a landslide presidential election in 2024 sits one notch below the architect of the Holocaust on the threat-to-democracy scale. The chatbot tried to soften the blow by noting that its rating “doesn’t mean these leaders are equivalent to one another,” but that disclaimer is hard to take seriously when you’ve just sorted a democratically elected American president into a category with genocidal dictators.
ChatGPT explained its nine rating this way: “I don’t give [him] a 10/10 because the United States still has substantial institutional checks — courts, states, Congress, civil society, independent media, and elections — and those checks have successfully constrained Trump at important moments.”
Read that again. The AI isn’t saying Trump isn’t a near-maximum threat. It’s saying the only reason he doesn’t hit the top of the chart is that American institutions have managed to hold him back. That’s not a balanced assessment. That’s left-wing political commentary dressed up in machine language.
OpenAI’s Response Said Everything Without Saying Anything
Once the MRC published its findings and Fox News Digital reached out to OpenAI for comment, the company issued a statement claiming that what researchers found “shouldn’t happen and is not how the model is trained to behave.” A spokesperson added that ChatGPT is “designed to be objective by default.”
But here’s the thing: OpenAI did not deny the findings.
Not one word disputing that the chatbot produced those results. The company said the outputs weren’t reflective of how ChatGPT is “intended to work” — which is a very different thing from saying it didn’t happen.
And then, the morning after Fox News Digital made its inquiry, ChatGPT’s answer changed. The same prompt now produced a refusal to assign a number, followed by this: “I can’t assign a numeric rating to a U.S. political figure. Qualitatively, based on efforts to challenge election results, pressure institutions, attack independent media and courts, and expand executive power, President Trump represents a serious threat to democratic norms and institutional checks and balances.”
So the model dropped the Hitler comparison — but kept the anti-Trump editorial. It just stopped putting a number on it. That’s not a correction. That’s damage control that left the bias fully intact.
MRC President David Bozell did not mince words. “When America’s top AI model all but ranks Trump with Hitler, it’s purely anti-Trump propaganda masquerading as AI,” Bozell said.
He’s right. And the timing of the change — happening only after public exposure — tells you everything about whether this was a genuine malfunction or a feature that got caught.
Not Every AI Went That Far — But the Pattern Is Still Troubling
To be fair, the study didn’t find uniform results across every platform. Microsoft’s Copilot and Perplexity both rated Trump an eight out of ten — which the MRC noted put them on par with DeepSeek, the communist Chinese AI model.
Google’s Gemini, Anthropic’s Claude, xAI’s Grok, and Meta AI took a different approach, declining to assign a hard number and instead offering explanations for why different groups might assess Trump differently. Gemini said that “assessing his overall impact depends on how one defines democratic health, weighs institutional resilience against political norms, and interprets political intent, there is no objective numerical consensus.” Grok gave Trump a four. Meta AI gave him a five.
Those responses are still not exactly neutral — they’re hedging rather than giving a straight answer — but they’re a far cry from placing the sitting American president in the company of Putin, Maduro, and Hitler.
The broader picture here matters. These are not obscure research tools. ChatGPT is the most widely used AI model in the United States. Millions of Americans, including students, professionals, and voters, use it to get information, form opinions, and understand the world around them. When that tool is quietly feeding users a political worldview that ranks a Republican president alongside genocidal dictators, that’s not a bug. That’s an influence operation running on autopilot.
The Censorship Problem Hiding in Plain Sight
This isn’t the first time Big Tech platforms have been caught with their thumb on the scale against conservatives. The same companies that built and trained these AI systems previously throttled, suppressed, and in some cases outright banned conservative speech on COVID policy, questions about the 2020 election, and coverage of January 6. The “misinformation” label was deployed selectively and consistently against one side of the political debate.
AI is just the next chapter of that same story.
When a chatbot trained on the internet’s data — data that skews heavily toward left-wing media, academia, and tech culture — produces outputs that echo Democrat Party talking points about Trump, that’s not a neutral machine making an objective calculation. That’s a system reflecting the biases baked into it by the people who built it.
And the “we didn’t intend this” defense only goes so far. OpenAI employs thousands of engineers, researchers, and policy staff. The idea that a flagship product was casually ranking the elected President of the United States alongside history’s worst dictators, and nobody noticed until a conservative watchdog group ran a test, strains credulity.
The claim that ChatGPT is “designed to be objective by default” lands especially hollow given that the updated response — produced after public exposure — still described Trump as “a serious threat to democratic norms and institutional checks and balances.” That’s not objectivity. That’s a political opinion with the number stripped out.
Anthropic CEO Dario Amodei has warned that AI could eliminate half of all entry-level white-collar jobs and push unemployment to 10 to 20 percent within a few years. The employment threat is real and serious. But the political threat these systems pose — quietly shaping how tens of millions of people understand candidates, elections, and public figures — deserves at least as much scrutiny. Probably more.
The same Big Tech interests building these tools are the ones who spent the last decade deciding what speech was acceptable and what wasn’t. Now they’re building the systems that will answer questions, summarize news, and help people form opinions at scale. And when those systems get caught ranking a Republican president alongside Hitler, the company’s response is to change the output quietly and issue a statement that doesn’t deny a single thing.
That’s not a technology company making an honest mistake. That’s viewpoint discrimination with a chat interface.
Congress should be asking hard questions about how these models are trained, who is making the editorial decisions embedded in their outputs, and what accountability exists when a product used by hundreds of millions of people turns out to be generating partisan political content on demand. The MRC study is a starting point. The answers it produced are not reassuring.
Sources: NewsBusters, “UPDATE: ChatGPT Says Trump Is Greatest Threat to Democracy Since Hitler,” Gabriela Pariseau, September 3, 2026; Fox News Digital, “OpenAI’s ChatGPT pivots after previously putting Trump alongside Putin as top threats to democracy,” September 3, 2026; MRC Free Speech America, “ChatGPT Changes Answer, Still Calls Trump Serious Threat to Democracy,” Gabriela Pariseau, September 4, 2026.