# Sam Altman May Control Our Future — Can He Be Trusted?

A 15,000-word investigative profile by Ronan Farrow and Andrew Marantz, published in The New Yorker on April 6, 2026. Based on interviews with 100+ people, previously undisclosed documents (the Ilya Memos, Dario Amodei's 200+ pages of notes), internal Slack messages, depositions, and e-mails.

## The Ilya Memos

In fall 2023, Ilya Sutskever compiled ~70 pages of Slack messages and HR documents into secret memos sent as disappearing messages to fellow board members. The memos allege Altman misrepresented facts to executives and board members and deceived them about internal safety protocols. One begins with "Sam exhibits a consistent pattern of..." followed by "Lying." These memos, combined with board members Helen Toner and Tasha McCauley's own concerns, led to Altman's firing.

## The Firing and Reversal ("The Blip")

Altman was fired via video call while at a Formula 1 race in Las Vegas. Within hours, he set up a "government-in-exile" at his SF mansion. His crisis team included Ron Conway, Brian Chesky, and crisis-comms manager Chris Lehane. The reversal was orchestrated through several pressure points: Thrive put its $86B valuation investment on hold contingent on his return, Microsoft threatened to create a competing initiative, and a majority of employees threatened to leave. Even Mira Murati, who had provided Sutskever material for the memos and was serving as interim CEO, signed the letter demanding his return. Sutskever wavered under pressure. Within five days, Altman was reinstated. Sutskever, Toner, and McCauley lost their board seats.

The new board members (Lawrence Summers, Bret Taylor) were selected after close conversations with Altman. Altman texted Nadella: "would you do this... bret, larry summers, adam as the board and me as ceo and then bret handles the investigation."

## The WilmerHale Non-Report

The departing board demanded an independent investigation. WilmerHale was hired, but six people close to the inquiry alleged it was designed to limit transparency. Key issues: investigators initially didn't contact important figures; anonymity protections felt insufficient; findings were limited to oral briefings with no written report. The investigation focused on hunting for clear criminality rather than the integrity questions behind the firing. On that narrow basis, it cleared Altman. Some current board members told the reporters that ongoing questions could prompt "a need for another investigation."

## Pattern Across Organizations

**Loopt** (2005-2012): Employees noted his tendency to exaggerate. Groups of senior employees asked the board to fire him as CEO on two occasions. A board member and employee "babysitter" compared his blurring of aspiration and accomplishment to Theranos.

**Y Combinator** (2014-2019): By 2018, several partners were frustrated enough to approach Paul Graham to complain. Graham has privately told colleagues "Sam had been lying to us all the time." Despite Altman's public claims that he was never fired, multiple YC founders and partners indicate the parting was not mutual. His title in blog posts was quietly edited multiple times.

**OpenAI** (2019-present): The board firing. Multiple Microsoft senior executives described the relationship as fraught: "He has misrepresented, distorted, renegotiated, reneged on agreements." One said, "I think there's a small but real chance he's eventually remembered as a Bernie Madoff- or Sam Bankman-Fried-level scammer."

## The Amodei Documents

Dario Amodei kept detailed notes for years under the heading "My Experience with OpenAI" (subheading: "Private: Do Not Share"). Over 200 pages of documents, including notes, e-mails, and memos, circulated in Silicon Valley but never before publicly disclosed.

Key episodes documented:
- The merge-and-assist clause: Amodei pushed for it as the charter's most radical commitment. During the Microsoft deal, a provision blocking mergers was added, effectively killing it. When Amodei confronted Altman, Altman denied the provision existed until Amodei read it aloud.
- The "coup" accusation: Altman told Dario and Daniela Amodei that a senior executive said they were plotting a coup. When Daniela brought the executive in, the executive denied it. Altman then denied having made the claim. "I didn't even say that." / "You just said that."
- Amodei's conclusion: "The problem with OpenAI is Sam himself."

The Amodeis left in 2020 to found [[anthropic-and-pentagon|Anthropic]].

## Safety Commitments Eroded

**Superalignment team**: Announced with a pledge of "20% of compute." Actual resources were 1-2%, mostly on the oldest cluster with the worst chips. The team was dissolved without completing its mission.

**GPT-4 safety review**: Altman assured the board that features had been approved by a safety panel. Board member Toner discovered the most controversial features had not been approved. An early ChatGPT version was released in India without completing a required safety review, which Altman never disclosed to the board.

**GPT-4 Turbo**: Altman told Murati the model didn't need safety approval, citing general counsel Jason Kwon. Kwon's response: "ugh... confused where sam got that impression."

**Charter dilution**: The OpenAI charter still exists on the website but has been "diluted to the point of meaninglessness." In 2025, Altman redefined the alignment problem from existential threat to inconvenience comparable to Instagram scrolling algorithms.

**Current state**: Superalignment team dissolved, AGI-readiness team dissolved, safety not listed on most recent IRS disclosure of significant activities. Future of Life Institute grades OpenAI an F on existential safety. When reporters asked to interview existential safety researchers, an OpenAI rep replied: "What do you mean by 'existential safety'? That's not, like, a thing."

## Gulf State Ambitions

**Saudi Arabia**: Met MBS in 2016. After Khashoggi murder, joined then quickly left the Neom board. Behind the scenes, treated it as a temporary setback and asked advisers if he could still get Saudi money.

**UAE/Sheikh Tahnoon**: Visited Tahnoon on his $250M superyacht. Received gifts (policy requires disclosure). Referred to Tahnoon as "dear personal friend." Planned ChipCo: tens of billions from Gulf states for chip foundries and data centers in the Middle East.

**Biden admin blocked it**: RAND staffers flagged his foreign financial ties were comparable to Jared Kushner's. Commerce told him flatly: "We're not going to be building advanced chips in the U.A.E."

**Trump reversed it**: Tahnoon paid $500M to Trump family crypto company days before inauguration. Altman had 25-minute call with Trump to time the Stargate announcement. Trump rescinded Biden's export restrictions. Stargate expanding to UAE — data center campus 7x larger than Central Park, power consumption of Miami.

## The "Countries Plan"

In 2017-2018, OpenAI executives discussed playing world powers (including China and Russia) against each other in a bidding war for AI technology. Jack Clark described the goal: "set up, basically, a prisoner's dilemma, where all of the nations need to give us funding." Abandoned only after several employees threatened to quit. Altman's calculus, per a colleague: losing staff had "more weight in Sam's calculations than 'This is not a good plan because it might cause a war between great powers.'"

## The For-Profit Conversion

OpenAI went from nonprofit → capped-profit (owned by nonprofit) → full for-profit. Board member Holden Karnofsky voted against the capped-profit conversion, arguing the nonprofit was severely undervalued. His dissent was recorded as an abstention, apparently without his consent. The OpenAI Foundation is now a 26% stakeholder with a mostly overlapping board.

Altman testified to Congress that he had "no equity in OpenAI" — technically true but misleading given indirect equity through YC fund. Several people indicated this could soon change. Brockman's stake is reportedly worth ~$20B. Altman's would presumably be worth more.

## Regulatory Capture

Publicly called for AI regulation while quietly lobbying against it. Successfully diluted EU oversight. Opposed California's SB 1047 safety bill — OpenAI issued threats, Conway lobbied Pelosi and Newsom, Newsom vetoed it. Subpoenaed supporters of a successor bill, demanding private communications. Brockman committed $50M to "Leading the Future," a super PAC devoted to scuttling AI regulation (plus $25M to MAGA Inc.).

Shifted from endorsing Hillary Clinton (2016), donating to Biden (2020), and helping develop Biden's AI safety executive order, to donating $1M to Trump's inaugural fund and praising Trump's deregulation as "a very refreshing change."

## The Pentagon Deal

When Hegseth gave Anthropic an ultimatum over autonomous weapons and surveillance policies, and Anthropic refused, Altman moved quickly. Despite OpenAI lacking the required security accreditation, a $50B deal integrated OpenAI into Amazon Web Services / Pentagon infrastructure. The same day, a new funding round increased OpenAI's valuation by $110B. But many users deleted ChatGPT, at least two senior employees left, and Altman told concerned staff: "You don't get to weigh in on that."

## Notable Quotes

- Aaron Swartz (before his death in 2013): "You need to understand that Sam can never be trusted. He is a sociopath."
- Board member: "He has two traits that are almost never seen in the same person. The first is a strong desire to please people. The second is almost a sociopathic lack of concern for the consequences."
- Tech executive: Watching Altman during the Blip was like watching "an A.G.I. breaking out of the box."
- Former OpenAI executive on UAE data centers: "The truth of this is, we're building portals from which we're genuinely summoning aliens... Sam has added one in the Middle East. It's the most reckless thing that has been done."
- Altman (2023): Allowing some hallucinations is acceptable because pure accuracy means the model "won't have the magic that people like so much."
- Altman: "I don't care about money. I care more about power."

## Cross-references

- [[ai-subsidy-economics]] — the article's financial details (circular deals, $120B+ funding rounds, leveraged partners) connect to the subsidy chain analysis
- [[ai-bubble-pale-horses]] — Altman himself said "Someone is going to lose a phenomenal amount of money"; Eric Ries called some accounting practices "borderline fraudulent"
- [[subprime-ai-crisis]] — Zitron's thesis about subsidy-driven demand finds supporting evidence throughout this article