Billionaire investor Stanley Druckenmiller has acknowledged using artificial intelligence to help write an opinion article criticising Treasury Secretary Scott Bessent, prompting a debate over whether wealthy public figures are held to a different standard from other writers.
The piece, published by The Wall Street Journal, attacked the US government’s decision to increase purchases of long-term Treasury bonds in an attempt to lower borrowing costs. Druckenmiller, a former mentor to Bessent, argued that the policy risked distracting from America’s wider spending and debt problems.
Suspicion about the article’s authorship grew after readers submitted it to Pangram, an AI-detection service, which indicated that much of the text had been generated by a machine. Critics described the writing as “Claudeslop” and questioned whether the Journal should have disclosed the use of AI before publication.
Druckenmiller did not apologise. He said he was not embarrassed by using the technology and rejected suggestions that the entire argument had been produced without his involvement.
The Journal’s opinion editor, Paul Gigot, also defended the decision to publish the article. The paper’s position was that the views expressed belonged to Druckenmiller, even if artificial intelligence had helped shape the wording.
The debate over AI-written opinion articles
The dispute has exposed a growing divide in attitudes towards AI-assisted writing. Editors and readers have often condemned undisclosed use of the technology by journalists, authors and academics, particularly when it has resulted in copied passages, fabricated quotations or factual errors.
Alex Preston admitted that sections of a book review published in The New York Times had been added using an AI tool after readers identified similarities with another review. The newspaper later added an editor’s note saying that the use breached its standards.
Other writers have faced professional consequences after being accused of relying on AI, even where no clear error was identified. Book deals involving novelists Mia Ballard and Jerry Falade were abandoned amid allegations that their manuscripts had been generated with artificial intelligence, while the literary magazine Granta stopped publishing winners of the Commonwealth Short Story Prize after criticism of its latest winner.
Those cases have frequently relied on detection tools such as Pangram. But such services do not establish authorship conclusively. Anthropic, the creator of Claude, has said its forthcoming text watermark can indicate only the likelihood that the system was involved; it cannot distinguish between text written by Claude and text that was heavily edited or proofread by it.
That uncertainty has intensified arguments for disclosure rather than retrospective investigations. Supporters of greater transparency say readers should be told whether AI was used to generate ideas, conduct research, edit prose or produce a finished draft.
Anthropic said in August that future versions of Claude would include an invisible watermark in generated text. The company has stressed that the system would not identify the individual user and would not prove that a person had used AI to write an entire piece.
The controversy surrounding Druckenmiller’s article has therefore become less about one op-ed on the bond market than about who is permitted to use AI without reputational damage. A billionaire financier may be judged primarily on the strength of his economic argument, while a student, journalist or novelist can face accusations of dishonesty for using similar tools.
The Journal’s decision to stand by the article is likely to add pressure on publishers to set clearer rules. Without consistent disclosure requirements, readers are left to rely on imperfect detection systems and online speculation to determine how published work was produced.
