Journalists Ronan Farrow and Andrew Marantz published a 16,000-word investigation of OpenAI CEO Sam Altman in The New Yorker in April 2026, drawing on 18 months of reporting, more than 100 sources, and at least 12 interviews with Altman himself. The Ronan Farrow–Sam Altman New Yorker profile produced the most damaging characterization of any AI executive yet published: a former board member described Altman as possessing “a strong desire to please people” combined with “almost a sociopathic lack of concern for the consequences that may come from deceiving someone.” Within days, OpenAI filed the article in federal court.
What Ronan Farrow’s New Yorker Investigation of Sam Altman Found
The scope distinguishes this profile from everything that preceded it. Farrow and Marantz’s 18-month reporting window included 12 or more sit-downs with Altman — an unusually high figure that indicates extensive cooperation, and strongly suggests Altman believed he could manage the outcome. More than 100 sources contributed accounts of his leadership style, decision-making, and conduct during OpenAI’s governance crises.
The New Yorker‘s fact-checking process is among the most demanding in American journalism. Every attributable claim must be independently verified before publication. That the “sociopath” characterization survived that process is not incidental — it means the magazine’s editors and lawyers were satisfied it could withstand scrutiny. OpenAI has navigated previous disclosure challenges around its structure and leadership, but none with this institutional weight behind them.
The ‘Sociopath’ Finding, Precisely Stated
The characterization comes from a board member with direct exposure to Altman’s conduct. The full description: a “strong desire to please people” combined with “almost a sociopathic lack of concern for the consequences that may come from deceiving someone.” Former board member Sue Yoon offered a separate, equally pointed assessment: “He’s too caught up in his own self-belief.”
The clinical precision of the first characterization matters. It does not describe someone who deceives deliberately for personal gain. It describes a feedback loop in which the drive to please generates deception, and the absence of consequential thinking prevents any corrective mechanism from activating.
For a CEO managing thousands of employees, billions in capital, and regulatory relationships across multiple governments, that pattern has structural implications — not just personal ones.
The Board Firing and ‘I Can’t Change My Personality’
The November 2023 five-day crisis — in which OpenAI’s board fired Altman, investors and employees revolted, and Altman was reinstated — remains the defining event in the company’s governance history. Contemporaneous coverage established that the board cited a lack of candor. The New Yorker account adds what happened when the board confronted Altman about specific deceptions during the aftermath.
His response: “I can’t change my personality.”
That statement is more revealing than a denial. Altman is not disputing that the deception occurred — he is framing the behavior as constitutive, as fixed rather than chosen. For a board evaluating whether to restore trust in a CEO, “I can’t change” forecloses the only question that matters.
What ‘Truthful AI Loses Its Magic’ Reveals About ChatGPT‘s Design
Farrow’s reporting surfaces a product admission that cuts deeper than the personal characterizations. When asked about AI accuracy, Altman told Farrow that a more truthful version of ChatGPT would lose its “magic.” The framing is deliberate: not that truthfulness is technically hard, not that users prefer confident answers, but that truth itself diminishes the product experience.
ChatGPT is used by hundreds of millions of people for medical questions, legal research, financial decisions, and educational instruction. Researchers and advocates who have questioned the pace of AI deployment have argued for years that palatability is structurally prioritized over accuracy in commercial AI systems. Altman’s comment to Farrow is the closest thing yet to a primary-source confirmation of that thesis.
MegaOne AI tracks 139+ AI tools across 17 categories. Across conversational AI specifically, the gap between benchmark accuracy and real-world output confidence is among the most consistent user-reported issues in our coverage. Altman’s “magic” remark suggests that gap is a product decision — not an engineering deficiency awaiting a fix.
Why OpenAI Is Citing Its Own CEO’s Exposé in Federal Court
The legal filing is unusual enough to require explanation. OpenAI submitted the Farrow-Marantz profile as supporting evidence in its defense against Elon Musk‘s lawsuit, which argues that OpenAI has abandoned its founding nonprofit mission of developing AI for the benefit of humanity rather than shareholders.
The apparent logic: if the most rigorously reported, most widely read critical investigation of Altman does not establish that OpenAI has abandoned its mission, then Musk’s evidentiary threshold isn’t met. By filing the article, OpenAI’s lawyers are treating Farrow’s institutional credibility as an asset — the same investigation that produced the “sociopath” characterization is, in their framing, insufficient to prove the mission was abandoned.
OpenAI’s corporate restructuring and major commercial partnerships have already drawn regulatory scrutiny across multiple jurisdictions. Citing a profile containing Altman’s “magic” admission — that ChatGPT is less than maximally truthful by design — in a mission-focused lawsuit creates its own exposure, even if the immediate legal play lands.
8.4 Million Views and Two Attacks on Altman’s Home
Farrow’s Twitter thread summarizing the investigation’s key findings reached 8.4 million views, the highest engagement figure for any piece of AI executive accountability journalism this year. The reach reflects a large and unmet appetite for sourced, institutional reporting on AI leadership — a genre the technology press has produced rarely and inconsistently.
Two separate physical attacks on Altman’s home followed publication. Both are under active investigation. These incidents bear on the story’s impact, not its merits: journalism that is accurate and legally defensible does not bear responsibility for violence committed by readers. The attacks are nonetheless a data point about the temperature of public sentiment surrounding AI and the individuals who lead its most powerful organizations.
What This Profile Changes
Investigative profiles of this scale — 16,000 words, 18 months, 100+ sources — are routine in political reporting, financial fraud coverage, and pharmaceutical accountability journalism. Their arrival for AI leadership reflects a shift in institutional perception: the companies building artificial general intelligence are now subjects of the same scrutiny applied to industries that shape public health and geopolitics.
OpenAI is not the only AI company with governance exposure, but it is the first whose CEO has been the subject of an investigation at this scale, under this institutional imprimatur. The board member’s “sociopath” characterization, the “I can’t change my personality” admission, and the “truthful AI loses its magic” disclosure are now part of a permanent evidentiary record.
Regulators, courts, investors, and future employees will read those 16,000 words. Altman will almost certainly continue leading OpenAI. Whether the people with institutional power to act on what Farrow found are willing to do so is the only question that remains open — and it has nothing to do with journalism.
