This updates me to 90-10 fraud. Sophisticated parties generally don't plead guilty to crimes they did not commit, and both Ellison and Wang are well-represented. I was probably overly optimistic with earlier predictions, perhaps in part due to positive personal interactions with Sam. 

Very sad for everyone involved. 

These are very sound, reasonable points. Agreed that the allegation that he repeatedly raised Alameda's line of credit, and its ability to carry negative balances, seems troubling and inconsistent with his claims that, "I had no idea." But complaints always sound bad. Hard to know what to make of the story until we've heard the other side. 

I also would feel more convinced if they stated in the complaint something like the following, "When other employees raised concerns that turning off risk controls for Alameda would be unfair and fraudulent, Bankman-Fried responded, 'I don't care if it's dishonest. We need to make money.' " 

Complaints don't have to go into that level of detail, but given the public nature of this prosecution, if I were the USAtty, I would have chosen to do so. The other side will get this evidence eventually, anyways, because it's legally required discovery. And it will build more public confidence in the prosecution -- that it is a legitimate prosecution rather than mob mentality -- if they show that they have clear evidence of fraud. 

If they have evidence of this sort and are hiding the ball, the best explanation to me is that they are trying to soft pedal their witnesses, to prevent them from backing out of their commitment to provide testimony. 

PS I'm curious. What was your take? I have a defendant bias, as someone who has done criminal defense work. Do you think I'm missing some smoking gun in this complaint that, at this point, shows that the prosecution has a very strong case? Fraud cases are very hard to prove, in my experience, even pursuant to the civil burden of proof. 

I read it. (It's here.) 

Doesn't really change my views that much. Maybe I'm 45 - 55 for mistake-fraud now, from 50-50. There's no smoking gun documents or communications in the complaint. It's possible they are hiding the ball, at this point. Perhaps they have a fragile witness who they don't want to expose to public scrutiny. (Caroline Ellison or one of the other high level executives seem like the most likely candidates.) But I don't see proof  beyond a reasonable doubt that this is fraud. 

The key thing to note is that, legally, Sam's subjective intent (and not whether customers were harmed or felt deceived)  is what ultimately matters. I think it also matters morally, though of course extreme negligence is immoral too. There are a lot of troubling circumstantial allegations made -- e.g., the fact that customer deposits were in some cases going directly into Alameda-controlled accounts, while Sam was promising folks that the two entities were separate and that customer accounts were segregated -- but unless the prosecution can prove beyond a reasonable doubt that Sam intentionally committed fraud, they should lose their case. (Does not mean they actually WILL lose their case, as in cases such as this, with so much public pressure, juries are quick to condemn even if the evidence is insufficient.) I can see an early-stage startup putting funds into the wrong bank account simply because it was convenient, and then forgetting about it later when it was convenient to forget about it. That's not fraud. That's negligence and incompetence. 

I still think the very fact that the feds filed charges is significant evidence that fraud was committed. But without hearing Sam's side of the story, it's really hard to say, from these relatively-barebones complaints, that there is clear evidence of fraud. Sam will likely have explanations for all the decisions made, e.g., the special coding for Alameda's margin ("We felt Alameda was a safe counterparty, and important to the business, and would have given the same treatment to any similarly-situated customer."). Until I hear that evidence -- or see CLEAR evidence of fraud --  it's hard to assess the relatively general allegations made in this complaint. 

An example of clear evidence of fraud would be something like this in the coplaint: "In August 2022, Bankman-Fried, realizing that Alameda's creditors were demanding more collateral, wrote to his team members: 'I don't care what the terms of service say, and I don't care what we promised investors and customers. Send the money to Alameda immediately, so we can solve their balance sheet problem. And don't tell anyone about this.' "

Instead this complaint has reasonable circumstantial evidence of guilt but nothing that, to me, leaps out as clear and convincing evidence of fraud. Of course, the prosecution will have much more time to develop their case. But if the complaint is all they have, I think they are going to have trouble winning this case. (I should add; the complaint is rarely all a prosecutor has, and I still think that SDNY probably would not have brought this case without something strong, e.g., a cooperating witness who can testify about agreements to lie to investors. Remembering the base rate of conviction, 90+%, is pretty important here.) 

I think it's very very fast. A few possibilities: 

1. Concerned he's a flight risk. 
2. Overwhelmingly clear evidence, most likely in the form of a cooperating witness who can provide documents proving fraud, e.g., records showing Sam saying, "We're going to have to lie to our customers to cover our losses." 
3. Concerned about additional harm he could cause, e.g., the mysterious withdrawals from FTX accounts shortly after the collapse. 

Of these 3, the last 2 seem most likely to me. Sam probably isn't a flight risk, given that he has deep connections in the United States. 

This updates me significantly in favor of thinking there is fraud. The conviction rate for the feds is 90+%, and the SDNY has a particularly good reputation – probably the best of any US Atty office in the nation. 

It's also possible that SDNY has jumped the gun due to public pressure. But I'm doubtful the US Attorney would bring this case unless there's pretty clear evidence of fraud. The US Attorney in SDNY, Damian Williams, is someone who has extensive experience litigating securities cases and presumably has personally vetted the evidence, given the prominence of this case. 

Very sad if it's truly the case that this is fraud. 

I think I've gone from 70-30 to 50-50, for the probability of mistake vs fraud. 

This is helpful. I might give the interview another listen with these particular issues in mind. 

What do we think might have been evasive in his answers? 

I would have to re-listen to the interview to see what mistakes he admitted. But I thought it was pretty clear what mistakes he admitted to, personally: failure to manage risk (e.g., understanding how likely it would be that there would be a simultaneous drop in collateral value, and a sudden withdrawal of deposits); failure to maintain corporate controls (e.g., creating red flags or account pauses, if certain accountholders like Alameda exceeded lending limits); and improper account segregation (admitted he did not know about commingling of funds). He did not admit to specific fraud, but that's consistent with the notion that he did not commit to fraud. He did admit to a number of things that, to me, provide a completely plausible account as to how FTX failed. And they are errors of risk management and even arrogance - but not malice or deceit. 

I've worked to pitch (and in some cases, been the target of) investigative pieces for the last 15 or so years of my life, and honestly, nothing here strikes me as particularly troubling. These are routine errors in communication, or cognitive biases (e.g., salience bias), and probably not indications of any sort of wrongful conduct. 

• Misstatements about Sam's frugality. It's possible there was an effort to mislead, but seems more plausible that this was just salience bias. A billionaire driving a Corolla is salient; owning a luxury condo in the Bahamas is not. Unless there is evidence that people actively misled the reporter, this is not particularly notable to me, other than reminding us all not to fall victim to that bias. 
• Warnings about Sam's misbehavior. Say someone causes others to find them completely ethical in 99.9% of all interactions. That's a very high rate! But once one becomes prominent, even a very high rate of perceived ethical behavior will lead to a high number of warnings because the number of interactions increases exponentially. Every prominent person has at least some people saying, "They're unethical." This is often because the prominent person merely turned a request down, when many requests are being made. (I am much less prominent than Sam but have had this happen to me many times.) I find the failure to respond to the exceedingly vague allegations against Sam unremarkable. 
• Sam's  contradictions on malaria nets.  You can read this as a lie. You can also read this as changing sentiments. We all contradict ourselves, and this is especially true when we are talking about highly speculative questions, such as cause prioritization, where our evidentiary basis is often quite slim, and where much depends on relatively small differences in the assumptions we make (e.g., a small change in the probability of hostile AGI). It is possible Sam is lying about his commitment to malaria nets to cover up his crimes. It's also possible that he just changed his mind, or at different moments, has a different emotional and rhetorical commitment to various causes. To give another example, Sam once stated that he was very committed to animal protection. Over time, he shifted his commitments and seemed more focused on concerns such as AI. I don't see that as a lie, even though I disagreed with it. It's just change. 

The main thing that I would find concerning in this piece is the excessive focus on PR by EA leaders. Don't focus on PR. Focus on trying to get a true and accurate account out there in the media. It's very hard to manipulate or even strategize about how to portray yourself. It's much easier to be real, because you don't have to constantly perform. That should be a norm within EA, especially among leaders. 

Great post and glad to see contrarian takes. (That's true as a general matter but I also happen to agree with this one :P)

Couple quick thoughts: 

1. Loyalty is important not just as a personal virtue but for efforts at collective action because it convinces people to engage in long term altruistic thinking. Hahrie Han has done some important empirical work showing that people are motivated to help when they have a sense of a shared past, shared future. Sudden ruptures in social relationships are very destructive for fostering that culture. 

2. Good decision-making generally does not involve dramatic changes to our beliefs. This is one of the less-known aspects of Tetlock's research on super predictors. They very rarely make large updates but are rather making small updates based on continuous data; they don't overcorrect when the mob moves. It seems likely to me that this is an example where we can try to put Tetlock's research to effect. I don't see the sort of dramatic evidence I would need to change my mind about SBF (though I also probably did not view him as highly as many others; my prior was that Sam was a very smart guy who was well intentioned but going in the wrong direction in life.) 

3. Good decision-making requires avoiding the Fundamental Attribution Error. I imagine most people are aware of the FAE on this forum. But I blogged about systemic forces that seem more important to me, in the collapse of FTX, than any of Sam's personal misdeeds.