In San Francisco, it might be old news that you shouldn’t believe everything Sam Altman tells you. But word has now reached New York. Last week, The New Yorker published (opens in new tab) a 16,000-word story by Ronan Farrow and Andrew Marantz that asked: Is Sam Altman someone we can actually trust with the future?

Farrow and Marantz spent 18 months investigating to answer that question. They interviewed Altman a dozen times and spoke to more than 100 people in his orbit. They followed unconfirmed rumors down windy rabbit holes and experienced first-hand some of Altman’s notorious Jedi mind tricks. 

If you haven’t read the article yet, you should get on that, but we understand it’s a lot, especially as chatbot brain destroys our attention spans.

So we invited Farrow onto “Pacific Standard Time” to break down what he found — and what it means for those of us living in the same city as the man who may or may not be building the thing that ends the world as we know it. 

Here are six revelations from the discussion that stuck with us.

OpenAI is using Farrow’s story against Elon Musk in court. After the piece was published, Farrow says OpenAI sent a letter to two attorneys general saying they trusted the story’s depiction of how the company was founded and what happened after Altman was ousted and then reinstated. and are citing it in their defense against Musk’s suit against the company, which alleges that the company abandoned its original nonprofit mission.

When confronted about lying after he was fired, Altman reportedly told the board: “I can’t change my personality.” Farrow reports that this was Altman’s response on a call where the board was pressing him to acknowledge a pattern of deception and people pleasing. He later told Farrow he might have been talking about his ability to bring disparate factions together. 

Altman reportedly admitted that a more truthful AI would lose its “magic.” Farrow connected the dots between Altman’s well-documented people-pleasing tendencies, as well as his loose relationship with the truth, and the sycophancy and hallucinations baked into ChatGPT. 

It’s not just OpenAI. Farrow says Anthropic has also watered down its safety commitments. Many people in this city treat Anthropic as the responsible adult in the room. Farrow’s view, after 18 months in the world of AI: The race-to-the-bottom dynamic is industrywide.

Farrow is convinced the AI risks are not coming — they’re already here. Farrow is less worried about a future AI apocalypse than he is about the real-world disruptions that are already underway. He’s talking about economic disruption happening now, AI’s documented role in mental health crises, and the technology going rogue. “It is not a projection,” he said. “It is happening.”

Growing up under intense public scrutiny informs how Farrow reports on power. Farrow and I discuss one thing we have in common: famous parents. My father is actor Richard Dreyfuss, while Farrow is the son of Mia Farrow and Woody Allen. He talks about how the “bad-faith” reporting his family has endured over the years helps him connect with sources who find themselves under the intense gaze of the media.

Watch the full interview here or on YouTube or listen to “PST” wherever you get your podcasts.

Read a lightly edited and condensed version of our conversation below.

Emily Dreyfuss: You recently published a bombshell story in The New Yorker with your coauthor Andrew Marantz about Sam Altman. After your story was published last week, there were two attacks on Altman’s home here in San Francisco. 

Earlier in the week, when Altman responded to your piece, he called it incendiary and said it could make things more dangerous for him. Do you think that’s fair? Are we at a point where people are just so fed up with these AI barons that they’re taking their anger to the streets?

Ronan Farrow: Absolutely, there is a moment of frustration nationally about a perceived crisis of legitimacy in AI leadership. That is due to legitimate questions about the legitimacy of leadership and the lack of guardrails on an industry that, by the admission of key people in it, could affect all of our futures so acutely, that has implications for warfare, chemical weaponry, disinformation, the economy, and that by even the most conservative economic projections could threaten millions and millions of jobs.

Understandably, it’s reaching the kind of breaking point that you’re talking about. It was before this particular story. It goes without saying that there should never be threats of vandalism or violence about any of the figures involved in this. I want all of these people to be safe. I want Sam Altman and his family to, you know, not at all feel fearful. 

It’s important to note this story was meticulously fact-checked, and OpenAI is not disputing the core facts. The reporting here is reliable. Critical journalism is not to blame for a moment of national anxiety like this or for these kinds of attacks. It is an incorrect assertion to draw any causal link between this reporting and the throwing of the bottle.

There is ample polling showing that a majority of Americans now have more questions and more perception of downsides about this technology than upsides. 

That is grounded in the very serious risks of this technology. Risks, by the way, which Sam Altman was one of the biggest alarmists about. The “this may kill us all” language; “this is more powerful than nukes” language. The acknowledgement about jobs being lost. 

When we talk about tamping down on anxieties, the fact is that the basis for those anxieties exists in truth around the technology. And when it comes to the question of how doomy is the rhetoric and is that alarming people, Sam Altman has built a company on that kind of doomy rhetoric. 

Let’s talk about what you learned in your 18 months reporting this story. The takeaway a lot of people have had from your piece is that Sam Altman appears to be a liar. Would you say that’s a fair assessment?

We talked to so many people and reviewed so many internal documents where there were complaints and alarms about this alleged trait. We carry and explore a range of perspectives on the question you’re asking.

There are parties that we talk to who say, well, “This is just so much compulsive lying that it’s dysfunctional for any executive of a major company.”

There are the people who say that is too much of a liability just from a business standpoint. And I’m talking about major investors, people with a stake in OpenAI, people connected to Altman who want to see him succeed and are concerned. 

Then there’s the opposite end of the spectrum, where people say this is a technology that requires a unique level of integrity. And this alleged trait of serial deception is not just untenable practically, but also actively dangerous.

So we look at both of those sets of views with a lot of scrutiny, and we also carry the views from defenders, anyone Sam himself wanted us to talk to and hear out. And I think the reason the core facts have not been in dispute is it is a genuinely fair and measured piece of reporting. The reader can decide on the answer on the question of trustworthiness themselves.

What drew you to Sam Altman as a character?

I had just done a large body of reporting on Elon Musk, and what emerged in that piece was a picture of tycoons in the current era, particularly in Silicon Valley, acquiring super governmental power. In Elon’s case it was through his Starlink technology. He was literally making material decisions for the outcome of the war in Ukraine and shutting down communications on the front lines. So that was fascinating to me. I went into looking at AI as the absolute outermost extreme edge of that imbalance.

My approach to this was not, you know, that Sam’s lying on an individual basis is so very alarming to me. It’s because it represents a wider crisis of integrity in leadership of this technology and OpenAI’s dealings with the political systems around them.

OpenAI was created as a nonprofit that was supposed to make sure safe AI was available for everyone. And somewhere along the way it seems to have become a company that just cares about money and power. Is it now just very blatantly true that OpenAI has kind of given up those earlier idealistic goals?

The fundraising pitch initially was, “This technology is more dangerous than nuclear weapons.” We have Sam saying that to Elon Musk. And Elon was very much a doomer at that point in time himself. And it was about appealing to that sensibility.

Since the development of consumer-facing chatbots and the blowing up of AI, it has become a race which was one of the things that OpenAI in its initial commitments was committed to preventing.

We have a moment in the piece where an internal whistleblower, a guy named Jan Leike, emails the board and says, “OpenAI is going off the rails on its mission” and talks about the risk of a race to the bottom, saying the problem is not just OpenAI. This is teaching Google and other competitors that they too should cut corners and race.

For your reporting, you spoke to Sam Altman numerous times, and he’s described [in the piece] as a master manipulator by some. What was it like to interview him? Did you feel at any point like he was trying to use Jedi mind tricks on you?

Certainly. Sam Altman’s traits, that so many people talk about in the piece, were at times in evidence in the extensive conversations I was having with him. I think I came to understand Sam somewhat deeply. I have tremendous sympathy for Sam in many ways. And I feel a great burden of, you know, fairness, and to do right by Sam —

— and where does your feeling of sympathy come from?

Sam’s explanation for the lying is that especially earlier in his career, he had a conflict-diversion and people-pleasing tendency that he acknowledges has caused him problems. 

I want to believe in the ability of people to grow and change. I want, when I’m reporting on someone like that, to give them the benefit of the doubt and be as generous as possible. I told Sam from the beginning, very transparently, “This is a piece that is scrutinizing some of the toughest criticism out there about you. It’s not gonna be fun to read.” So he participated knowing that.

Did he have to be convinced? Was he hesitant?

I’m not gonna speak for his state of mind beyond, you know, the things that I’ve quoted in the piece. I will say that anytime you have a body of reporting like this, it’s very pressurized and can be very combative and contentious. Very often there’s a lot of threats flying back and forth.

So it’s a draining, careful process where you just have to deal with the volatility that can be intrinsic to these moments. However contentious conversations may get, Sam Altman did choose to participate a lot — again and again. I am ultimately grateful because when you cut through any hostilities, what you get is someone revealing more about their point of view.

I had conversations where I was very directly trying to break through on a person-to-person level. You know, “What is your individual experience of everyone around you saying this?” You know, that would be devastating for me. “How do you carry that? Do you talk about that in therapy?” So I really did try to come to a place of maximal understanding, which hopefully also informs the very sober, even tone of the piece.

That actually brings me to something I wanted to talk to you about: your status as a child of Hollywood. I’ve been dying to ask you about this for years because my dad is the famous actor Richard Dreyfuss. This might be news to anyone who’s listening to the podcast —

Yeah, I didn’t know that.

Well, I’m very private about it. But when I watched your landmark reporting on the #MeToo movement and the way in which your background seemed to inform the way you were able to connect with the people in the story, it was really illustrative for me. I thought, here you are being so authentic in your experience.

How does your experience of power and fame help you to understand powerful people and maybe have empathy for them in these moments?

That is one of the core takeaways from my painful experiences and my dirty laundry in my family being so painfully on display. I have been targeted by bad-faith press, particularly when I did reporting that ultimately contributed to the National Inquirer and its parent company getting hauled in front of prosecutors right around the Trump hush payments. I was really getting beat up in the most bad-faith way possible. It’s scary. And all you can really count on is the reporting that you do being good faith and being the kind that you would want to receive. So I know the difference between those two things.

There are ways in which I’m in a position of luxury. I can pick my topics, and I can spend the time. I am immensely fortunate, but also it is harder in some ways. I’m always a lightning rod. I’m always an easy target. When people I report on are angry about [something], it is much easier for them to make it personal. The best defense I have is to just make sure the reporting is unimpeachably in good faith and fair.

Your reporting paints this picture of Sam Altman who is loose with the truth. We also hear about so many people who’ve known that forever. I mean, you have Aaron Swartz raising this in 2013.

What does this tell us about the culture right now in Silicon Valley? Are we dealing with a culture that is OK with a loose relationship to the truth?

Absolutely. What’s interesting about this piece is you have this baseline understanding that, even in the eyes of Sam’s defenders and even in the account of Sam Altman himself, this [lying] has been a problem. And the reaction around it is not like, well, “That’s untenable for people in the most powerful position, shaping the most powerful and dangerous technology ever.”

It is instead, well, “Isn’t that what executives in Silicon Valley do? Isn’t that being a founder? Isn’t that what it takes to build a startup?”

There’s a call we report on after [Altman] is fired and the board is pressing him to acknowledge a pattern of deception. And one of his responses, according to people on the call, was, “I can’t change my personality.” And in the present day he said, “I don’t quite remember [but] I might have been talking about my ability to be a peacemaker and tell the different factions the things that are needed.” I’m paraphrasing here, but the things that are needed to bring disparate groups together and build an enterprise.

And there is an understanding and embrace of a culture of hype. The whole story of Silicon Valley is in massively inflated valuations long before there are actionable deliverables and products. That has become so normalized. 

It’s easy to dismiss what’s happening in AI as part of that hype machine, but then the question is whether the worries get dismissed too, because the worries are also being hyped. And I think maybe what your reporting brings up is the fact that we’re at this moment where we actually need to face these worries head on.

I think that’s exactly right, and the piece fairly explores and presents the range of views. The safety-ists who are like, “Yes, this continues to be real, this could kill us all eventually.” The fact that we’ve already had verified, peer-reviewed literature on how fast these models can generate chemical weapon agents. The list goes on and on of economic, practical and safety concerns that have arrived.

We talk about the rift with [Anthropic founder] Dario Amodei and the way Amodei, and a bunch of others, left OpenAI because they didn’t trust Sam and they didn’t trust the development of the technology and how it was proceeding under OpenAI. They wanted to preserve that original safety mission. We acknowledge in the piece that Anthropic has, in some respects, more material on safety commitments. But we have also seen Anthropic water down many of their key safety commitments. And we see them participating in some ways in that same, alleged critics say, race to the bottom. So there’s just shrinking space for anyone who believes in the risks and wants to structure a business mainly about that. 

I — for what it’s worth — emerged from the piece fairly convinced, whether you believe the far-out scenarios or not, of the risks that are materializing right now to jobs, to our safety, to the economy —

—And not just the economy, right? I mean, it’s also the social and psychological stuff where people are relying on these chatbots to cure their loneliness and not calling their friends. I mean, these are the things that are happening now.

Deeply. A great example is the twin problems of sycophancy and hallucination in these AI models. Sycophancy, being the way they suck up and they mirror back whatever the user believes and wants to hear. Hallucination being when they’re generating answers and predicting the next word that should go next. There’s this strong imperative to sometimes make up things. 

So you have human-reinforced training that these labs are using where they know at this point that the problem of hallucination has not been solved. And that human reinforced learning reinforces sycophancy, right? Humans like agreeable answers. So they rate those higher. So the models learn to say back the thing that’s gonna please the person or group in question most.

And it goes back to these broader structural issues. This is not about Sam Altman. This is about an industry without guardrails, without meaningful oversight, and a lot of stakes that are already in play for the rest of us.

We’d be remiss if we didn’t point out that these issues that you just brought up of sycophancy and hallucination remind me of the things people are saying are true of Sam Altman’s personality as well — that he’s a people pleaser who has a loose relationship with the truth.

Well, we quote Sam Altman saying that you could create — in his view — a model that was more factual, but that that would lack some of the magic that people love.

I think that speaks for itself.