Billionaire Investor Druckenmiller Admits to Using AI in WSJ Critique of Treasury Secretary
Billionaire investor Stanley Druckenmiller admitted that he used AI in his column criticizing Treasury Secretary Scott Bessent's bond buyback policy in The Wall Street Journal, stating that he now "uses AI for anything he writes."
The article, titled "Let the Bond Market Speak," opposes the U.S. Treasury's expansion of long-term bond buybacks to suppress yields. Druckenmiller believes that buybacks of $4 billion or more cannot address the fundamental causes of rising government financing costs, referring to it as "price management" rather than liquidity management.
The controversy arose when economist Claudia Sahm submitted the article to the Pangram detection tool, which indicated that the text was "100% AI." This output only reflects the tool's probabilistic judgment of text patterns and is insufficient to prove that the article was entirely generated by AI; subsequently, NOTUS reporter Jeff Stein directly confirmed with Druckenmiller, who acknowledged using AI.
Druckenmiller denied that "the entire article was generated by AI." He stated that he rejected many AI suggestions, asserting that the article expresses his views and likening AI to a calculator: a tool to assist in writing without altering the author's responsibility for the views and conclusions.
Paul Gigot, editor of the editorial page at The Wall Street Journal, defended the publication of the article, stating that the core judgment of submissions is whether they represent the author's original argument and whether the author has the qualifications and credibility to express that viewpoint; he noted that Druckenmiller's article reflects his genuine position.
In market mechanisms, AI writing reduces the marginal costs of organizing language, summarizing information, generating drafts, and editing to nearly zero, but investment judgment, factual responsibility, authorship credibility, and legal risks remain with humans. Model suppliers and enterprise software platforms will benefit from default usage; services focused on standardized copy production will face pressure, while those able to provide unique insights, original data, industry relationships, and accountable decisions will command a premium.
Source: Public Information
ABAB AI Insight
Druckenmiller's writing habits align with his tool perspective throughout his investment career: he transitioned from an English major to economics, later establishing a macro trading reputation within George Soros's framework, emphasizing macro variables, position management, and risk discipline through Duquesne Capital. This bond article targets not an abstract fiscal topic but the specific policy advocated by his former subordinate Bessent to expand bond buybacks; in this high-risk public expression, he used AI for language production while retaining judgment, consistent with the tradition of macro investors separating computation, information processing, and final bets. In terms of capital pathways, AI first enters the "expression layer" of high-income decision-makers: emails, memos, commentary articles, initial drafts of investment frameworks, and meeting preparations can be quickly generated, rewritten, and condensed. What it saves is not the viewpoint itself but the time to convert the viewpoint into distributable text. Model companies profit from high-frequency reasoning calls, and media benefit from more prolific submissions; however, when AI-assisted texts are not transparently disclosed, new trust costs will emerge between editorial departments, readers, and authors, especially in fields like markets, policy, and healthcare that may influence decisions. Historically, calculators did not eliminate mathematicians, Excel did not eliminate investment analysts, and Bloomberg terminals did not eliminate fund managers; they shifted competition from manual calculations to variable selection, interpretive ability, and position decisions. The role of generative AI in writing is similar: it automates rhetoric, structure, and revisions but cannot replace verifiable judgments on fiscal deficits, long-term interest rates, bond supply and demand, and policy consequences. The "100%" result from Pangram also illustrates the limitations of AI detectors—they identify text styles but do not equate to confirming the complete creative process. Essentially, this is a matter of technological substitution. The first to be substituted are the manual labor tasks between the ideas in the mind and formal text, including drafting, rewriting, formatting, and tone adjustments; what is not substituted is reputational capital. The authorship of high-impact authors still means their viewpoints, conflicts of interest, and error costs can be held accountable. The mechanism of change is that language models commodify general expressive capabilities, while scarce resources shift to "who possesses genuine judgment, exclusive information, and the identity to bear consequences." ABAB News · Cognitive Laws
- AI can replace writing but cannot replace authorship responsibility.
- Expression can be outsourced, but judgment must be borne by oneself.
- When text can be generated infinitely, credibility becomes a scarce asset.