Stanley Druckenmiller, chairman and CEO of Duquesne Family Office, did something increasingly common and unusually revealing: he used AI to write an opinion column criticizing Treasury Secretary Scott Bessent.
Druckenmiller later said he uses AI for nearly everything he writes and compared the tool to a calculator. He maintained that the opinions were entirely his.
From Nieman Lab:
Any good op-ed can spark a public conversation, but most won’t start a debate over the nature of opinion writing itself. Last week, a Wall Street Journal piece by the billionaire financier Stanley Druckenmiller pulled it off.
The op-ed itself is about whether the Treasury Department should continue trying to lower long-term bond yields. (Druckenmiller says no.) The piece has a few telltale signs of being produced by AI. (“This wasn’t liquidity management, it was price management.” “If the 30-year must trade at 5.5% to clear, that isn’t a crisis. It is an invoice.”) NOTUS reporter Jeff Stein rang him up to ask and Druckenmiller acknowledged it.
Now, in most cases of that-smells-like-AI discourse, this is the moment where the writer offers some combination of denial, explanation, and apology — and the piece gets an editor’s note or taken down altogether. What made this episode different is that neither Druckenmiller nor Journal editorial page editor Paul Gigot offered any regrets or shame. “I write everything using AI now for the same reason I use a calculator when I do math problems,” Druckenmiller said.
“AI is a fact of modern life…nobody can doubt that his op-ed is his genuine opinion,” Gigot said. In a follow-up column, Gigot said there was no need to disclose Druckenmiller’s AI use, arguing that “it wouldn’t be honest for me to grandstand that our contributors never use a tool like AI. Mr. Druckenmiller’s sin against the media pharisees was admitting its use.”
AI can produce competent sentences. The harder question is what readers believe they're getting when a human name sits above an opinion piece. Are they buying only the conclusion, or are they paying to encounter someone's judgment, language, personality, and thought process?
One editorial page editor, the San Francisco Chronicle’s Matthew Fleischer, put it almost perfectly: "Our audience is paying for the words and thoughts of human beings."
Nieman Labs reached out to editors at 24 prominent newspapers and asked about AI-written opinion pieces, and none of those who responded embraced the broad approach that allowed Druckenmiller's piece.
Several papers explicitly prohibit AI-generated op-eds.
The problem has already moved beyond theory. A review of 54 recent political op-eds in New York found 28 containing text flagged by an AI detector. More importantly, nine politicians acknowledged that they or a co-author used AI in producing 10 pieces.
From New York Focus:
If you’ve read an op-ed attributed to a New York politician in the last few months, there’s a good chance that not all of it was written by a human being.
To track the rising use of artificial intelligence in political communications, New York Focus analyzed 54 opinion articles attributed to New York elected officials that have been published since May 1 by three prominent news outlets: City & State, the Albany Times Union, and the New York Daily News. We ran each of those op-eds through the AI-detection tool Pangram and contacted the officials whose pieces were flagged as containing AI content.
Nine politicians — including the governor’s running mate, the mayor of Albany, and the chair of the Senate technology committee — acknowledged that they or a co-author used AI in producing 10 different articles. Another five said they could not be sure whether AI was used, four denied using AI, and nine did not respond to requests for comment. Twenty-six of the articles were classified by Pangram as fully human-written.
Just over half of the pieces — 28 in total — included text that Pangram identified as AI-generated.
None of the pieces disclosed the use of AI — even though lawmakers passed a bill this year that, if signed by the governor, would require news outlets to disclose substantial AI content. (One outlet, City & State, added disclosures to some articles after receiving New York Focus’s questions.)
The researchers wisely declined to accuse everyone flagged because detection software can make mistakes.
Used with discipline, AI may become one of the most valuable tools in the modern workplace. Manufacturers can turn engineering changes into draft work instructions. Construction firms can prepare safety documents and bids.
Logistics companies can create procedures and test cases. Sales departments can shape proposals. Marketing teams can produce campaign variations. Insurance offices can draft routine correspondence. Medical offices can reduce administrative paperwork.
Sales and marketing, software engineering, customer service, and research and development already represent some of the largest areas of potential economic value from generative AI.
Journalism has legitimate uses for it, too. Current newsroom standards at one major news organization permit AI for early research, document summaries, transcription, translation, headline suggestions, grammar, spelling, and search optimization. Human journalists remain responsible for verification, editorial judgment, and publication.
Opinion writing reaches a more personal boundary because voice is part of the product. Politicians and executives have employed speechwriters for generations, so delegated prose existed long before ChatGPT.
Human ghostwriters can question an argument, press somebody to explain himself, recognize evasion, and demand another draft.
A machine, however, can generate polished copy from five bullet points before the coffee gets cold.
An executive can approve 800 words of machine-generated prose that accurately represents his position. Readers still deserve to know who supplied the sentences.
Using AI for research, organization, transcription, fact-checking, or a grammar pass doesn't trouble me. Let the technology eliminate tedious work wherever it can, and when AI supplies a substantial share of published prose under a human byline, disclosure should follow.
And here's the part where I make my full disclosure. I use AI for research, and it's simplified my writing so much that I'm still having a hard time grasping the concept. Before writing for PJ Media, I had been writing for decades using several pens: ad copy, travel brochures, technical writing, marketing campaigns, and content marketing.
My previous two content writing jobs involved deciding on a topic, deep-dive research, formulating a summary, and then beginning writing.
To complete that strong piece of content, it took me many days. Now? I enter a search into ChatGPT, asking whether there are any statistics, deeper history, or anything else that pops into my melon.
It takes a bloody minute! It's a fantastic piece of technology that I don't take for granted.
That's the AI tool I use when writing for PJ Media, which has a rigid anti-AI policy — and that is the right thing. Fleischer's quote is 100% correct: you're not here to read something pieced together by an AI that's collected words for several years and dumped them into a bucket for a lazy writer. You're reading my goofy, disjointed thoughts that hopefully make sense.
A byline should identify more than the person who approved the final draft; it should identify the person whose thoughts and words the reader came to read.
Want more reporting and commentary without the usual media filters? Join PJ Media VIP today and save 60% with promo code FIGHT.
Join the conversation as a VIP Member