The most dangerous thing about artificial intelligence may not be what the machine thinks. It may be who gets to tell it what is true.
That fight burst into the open last week, when the Babylon Bee, a conservative Christian satire publication, sued New Mexico over a law it says requires AI-generated political satire to carry a government-mandated disclaimer. The lawsuit exposes something bigger than whether a joke needs a warning label: Who gets to classify speech before the rest of us get to judge it?
AI systems constantly classify information — reliable from unreliable, biased from neutral. Totalitarian governments have long exercised a similar power, deciding that certain categories of speech are suspect enough to require warning, restriction, or suppression. Different technology. Same impulse: control the category before the public gets to judge the claim.
COMEDY WITH A WARNING LABEL: NEW MEXICO WANTS THE BABYLON BEE TO SPOIL ITS OWN JOKES
For years, we’ve been warned that AI is biased. The fix sounded simple: “debias” it. One problem. Who decides what counts as bias? The moment somebody decides which answers need correcting, bias hasn’t vanished — someone has simply acquired the authority to define it. Machines inherit human blind spots from their training data. But when engineers or bureaucrats “correct” those judgments, they don’t become neutral. They substitute their own for the ones they dislike.
The referee didn’t disappear. We hired a new one — and called her objective.
I came to this realization as a trustee of the Wilson Center, watching experts help policymakers grapple with AI bias. The problem was real. What troubled me was the cure: Every attempt to “debias” a system required somebody first to decide what counted as bias.
I’ve tried to build toward a different model. I hold a U.S. patent on technology that validates AI data through multiple independent validators, based on the proposition that truth is safer when it doesn’t depend on a single gatekeeper.
The problem is older than AI. Three centuries before ChatGPT, Rabbi Moshe Chaim Luzzatto, the Ramchal, built much of his method around a simple idea: Clear thinking requires organizing categories correctly. Get the category wrong, and the reasoning can fail before the debate begins. Whoever names the category has a head start on winning.
A restriction on speech becomes “trust and safety.” New Mexico doesn’t call its requirement censorship. It calls it disclosure.
Same move.
Sometimes our would-be arbiters say the quiet part out loud. Former Secretary of State Hillary Clinton warned on CNN in October 2024 that if platforms don’t “moderate and monitor” content, “we lose total control.”
Who is “we”?
Rep. Alexandria Ocasio-Cortez (D-NY), days after the Jan. 6, 2021, Capitol riot, said: “We’re going to have to figure out how we rein in our media environment.”
Listen to the verb. Not challenge. Not persuade.
Rein in. Control.
That’s the part of President Donald Trump that his critics still don’t understand. “Fake news” has power because underneath it is something true: An institution’s authority isn’t proof it’s right.
For years, people were told to trust the experts, the agencies, the newsroom. Now add: trust the algorithm.
Trump’s answer, stripped of theatrics, is: prove it.
That instinct is now federal policy. Trump’s AI procurement order requires government systems to meet standards the White House calls “truth-seeking” and “ideological neutrality” — prioritizing accuracy and honest uncertainty over baked-in ideology.
Those labels should be judged by the same standard: whether the evidence and assumptions behind them can be examined and challenged. But there’s a real difference between “Trust us” and “Here’s the evidence — check it yourself.”
The administration brought a related fight to the Smithsonian, accusing it of dressing up ideology as neutral history. Critics have a fair worry: Maybe this White House just swaps in its own version.
That possibility is precisely why the principle has to apply universally. It doesn’t defeat the argument. It proves it.
No institution, including this administration, gets to be the final word on truth. The cure for institutional bias isn’t another institution claiming infallibility. It’s contestability: show assumptions, show evidence, show what’s uncertain, and let competing claims fight it out in daylight.
We will never engineer bias out of people or the machines they build. Pretending otherwise may be the great con of the technology age.
What we can do is make bias visible — compare sources, expose assumptions, make institutions show their work — and stop pretending the people deciding what’s “misinformation” are immune to the same prejudices as everybody else.
That’s the real divide in politics now. The establishment’s instinct is mediation: better experts, better gatekeepers. The populist wager is riskier — and more democratic: let people look for themselves.
Yes, they’ll get things wrong. Freedom was never a guarantee against being wrong. Self-government rests on something more radical: citizens get to examine evidence, distrust authority, and make up their own minds.
The cure for biased machines isn’t programming in the preferred biases of whoever holds the keyboard. The cure for citizens who reach the “wrong” conclusions isn’t treating them like broken algorithms needing retraining.
That’s why the Babylon Bee case matters. A state can call a warning label “disclosure.” An engineer can call a judgment “debiasing.” Changing the label doesn’t change the power underneath it.
We don’t need a Ministry of Truth — not in Washington, Silicon Valley, a museum, Santa Fe, or a chatbot.
HISTORY WITH A WARNING LABEL: TRUMP IS RIGHT ABOUT WHO TELLS AMERICA’S STORY
Truth doesn’t require a final referee. It requires something harder and more democratic: evidence that can be checked, assumptions that can be challenged, and citizens free enough to decide for themselves whom to believe.
The greatest threat is not that machines will think. It is that someone will be given the power to decide, once and for all, what thinking is allowed.
Alan Rechtschaffen serves as a trustee of the Woodrow Wilson International Center for Scholars, a congressionally chartered institution within the Smithsonian. He writes here in his personal capacity. The views expressed are his own.