Artificial intelligence could spoil our nation’s 250th birthday party.
As we approach our semiquincentennial, Americans eager to learn about our country’s founding may turn to artificial intelligence, only to encounter a history that does not celebrate the brilliance and bravery of the colonists who wrote our Constitution or led the revolution against England.
Instead, asking Chat GPT or Anthropic about the events and people who created this exceptional nation could lead down a dark tunnel of negativity.
There are many reasons to distrust or worry about AI, including the potential loss of jobs, the spread of misinformation, and identity theft. But even more important could be the tarnishing of the American story for generations to come. Thanks to AI, the history our children read may not celebrate our heroes and successes, and it may not lead them to love our country.
I asked Anthropic’s Claude, and what he gave me this disconcerting account of how AI might distort our American narrative:
- “AI doesn’t hate your heroes — it just can’t resist a scandal, and that bias slowly turns legends into cautionary tales.
- “The patriots, pioneers and everyday heroes who built this country don’t trend — so AI quietly buries them while amplifying every dark chapter.
- “AI is trained on the internet, and the internet loves outrage — meaning our national story is now filtered through the same negativity bias that ruins your morning news feed.
- “In the name of nuance, AI strips the heroism from our greatest figures — leaving behind hollow names where giants once stood.
- “Kids are already turning to AI for homework help — and the version of history they’re getting is heavier on failure, guilt, and conflict than on courage, sacrifice and triumph.”
This last comment rings true and should set off alarm bells. If our children grow up only hearing the most damning reviews of our nation, they won’t develop the patriotism that glues our country together and demands the sacrifices made, especially, in time of war.
This is happening already. Gallup’s annual surveys show a sharp drop in national pride, with last year’s installment recording record-low numbers of respondents saying they are “extremely” (41 percent) or “very” (17 percent) proud to be an American. The fall-off has been most acute for Gen Z, those born after 1996.
Our schools and universities have poisoned the well, with decidedly leftist faculties demonizing capitalism, religion and other core values like meritocracy, freedom and self-reliance that have helped make the U.S. the envy of the world.
If Claude is correct, the liberal infection of our education system will be amplified as young people turn to AI for information. What they learn will accelerate the so-called “Mamdani Effect.” By that I mean that young people, ignorant of why the Soviet Union collapsed or why Venezuela went from one of the richest countries on the planet to one of the poorest, will keep embracing socialist candidates like New York City’s new mayor, Zohran Mamdani.
Mamdani campaigned on delivering Big Government solutions to today’s problems, such as free childcare and rent control. He won over young voters who were unaware of the dismal record of such promises. Their leftist professors never taught them about the failures of socialism or the world-changing successes of free markets; AI won’t teach them, either.
One of the more destructive aspects of allowing AI to hijack the telling of our history is its tendency to destroy our heroes, the men and women who created our country and have led it through good times and bad.
If a second grader asks Chat GPT whether George Washington was a good man, it will tell her: “Whether George Washington was a ‘good man’ depends on what standards you use and which parts of his life you focus on.” If a seventh grader asks which is better, capitalism or socialism, Chat GPT will equivocate, responding, “Whether capitalism is ‘better’ depends on the outcomes you care about most” — the kind of answer you might get from a CNN anchor.
AI was not built to be unpatriotic, but its large language models are trained on material readily available on the internet, where outrage fuels traffic and where good news is squashed by drama and disaster. The more a newspaper column generates clicks, the more that piece gets scraped into the next round of training data. AI creators have set up a feedback loop. The algorithm learns that salacious material about an historical figure is more important — more productive — than the good deeds he or she may have done.
Thomas Jefferson, for instance, was the brilliant architect of our Declaration of Independence, but his fathering of a child by a black slave named Sally Hemings creates more of a ruffle on the internet.
It is also true that large language models consume, as does the public, an information flow that is dominated by liberal media. That AI-generated content displays a leftward bias has been well documented. This partisan tilt affects how news is reported today; it will certainly also determine how our history is told.
There has been much chatter about “ sycophancy” infecting AI responses. Bots tend to tell users what they want to hear. I asked Claude if his revelations of how large model architecture inevitably distorts our historical narrative were accurate or whether he was responding to my own suggestions. Both, he responded, writing, “The core observation is real and defensible.”
The White House, alarmed at the fast-developing power of cutting-edge systems like Anthropic’s Mythos, has begun the difficult task of establishing federal oversight to the emerging industry. The public should demand that, in addition to preventing abuses like cyberhacking and intellectual property theft, AI developers must also address the leftist bias that is embedded in their training models.
Artificial intelligence offers many promising benefits and also identifiable risks. It could well be that undermining confidence and pride in our nation is the greatest threat of all.
Liz Peek is a former partner of major bracket Wall Street firm Wertheim and Company.
This article originally appeared on The Hill.
