The Trial of Sam Bankman-Fried

Three Managerial Errors

The alternative account focuses on these three errors. First and most disastrously, the leadership of FTX and Alameda lost track of the outstanding balance of a debt that Alameda owed to FTX as the result of a payment processing arrangement they had entered into in early 2020, when FTX could not get its own bank accounts. That debt, which was not actively tracked by Sam once he stepped down as CEO of Alameda, ballooned from $4 billion in the summer of 2021 to $9 billion by mid-2022, as customer deposits greatly outstripped withdrawals.

Second, the CEO of Alameda, Caroline Ellison, failed to hedge against two successive crypto market crashes in the spring of 2022, which reduced the value of Alameda's crypto holdings by about 70–80 percent—dropping its net asset value from a positive $40 billion at the end of 2021 to a positive $10 billion by June 2022. Even after paying off all its debts, including to FTX, Alameda would still have had $10 billion in assets. But many of those assets were illiquid, meaning they couldn't be converted to cash on the spot. That left FTX exposed to a classic bank run.

The Run on the Bank

What befell FTX in November 2022 was a classic bank run, triggered by a hostile competitor's decision to liquidate its holdings of FTT, the FTX exchange token. When Binance's CZ announced his intention to sell, customers began withdrawing funds in massive volumes. In the 72 hours before bankruptcy was filed, roughly $8 billion was withdrawn from the exchange—an amount that would have rendered any fractional-reserve financial institution insolvent on the spot.

Sam was actively seeking emergency funding to cover the shortfall. He received firm commitments for between $6 and $8 billion by the morning of November 11th, the last of which came in 10 minutes after he had turned FTX over to John Ray, the newly appointed CEO, who then filed for bankruptcy without consulting Sam. Those commitments, had they been accepted, would have covered the shortfall entirely.

Rush to Judgment

The government offered two different theories of why the loans in question were fraudulent. The first—the primary focus of its case—was misappropriation (theft): the loans came from funds that the government claimed were customer property, not FTX property, and therefore could not be loaned to Alameda without customers' knowledge and consent. The second was misrepresentation: Sam, in various public statements, misled customers about the risks the loans imposed on them.

The government rushed to indict. Under a hard-to-defend rule developed in the Second Circuit, it is permitted to introduce evidence relevant to dropped charges even if those charges were later severed—on the theory that such evidence is relevant to the charges that were tried. The six dropped charges in Sam's case were not without consequence: they allowed the prosecution to paint a sweeping picture of criminality that bore little relation to the actual allegations at trial.

The conclusion appears to have been foregone the moment the government got a tip—which ultimately proved false—that there might be some irregularities at the FTX US exchange. In federal white-collar prosecutions, the government rarely loses at trial. It gets to pick the venue (here, the SDNY—arguably the most prosecution-friendly jurisdiction in the country), bring massive resources to bear, and control the charging decisions that set the frame for the jury's evaluation of the evidence.

The Role of Cooperating Witnesses

As the prosecution acknowledged in its sentencing letters recommending leniency for each of the three co-founders, it could never have convicted Sam without their testimony according to its own sentencing submissions. Each testified pursuant to a cooperation agreement that all but guaranteed minimal or no jail time in return—provided that the government, after the fact, judged their testimony to have significantly helped its case against Sam.

In the case of incriminating testimony obtained from cooperating witnesses, the truth is, at a minimum, contaminated. Any lawyer who has dealt with cooperating witnesses—as prosecutor, defense counsel, or judge—knows that they are unreliable. Incentivized testimony from cooperating witnesses who have been coached by the prosecution for months is the most unreliable of all. The jury was told this in general terms, but was given no concrete tools to evaluate the specific incentives at work here.

A. The Misappropriation Theory

The government's view: even if the agreements were complex, the core issue was practical control: customer deposits were supposed to be available on demand, while Alameda was allegedly given hidden exemptions and lines of credit that ordinary customers did not have. On that view, contractual ambiguity does not authorize secretly subordinating customer withdrawal rights to Alameda's borrowing.

The government's misappropriation theory rests on the allegation that loans to Alameda violated customers' rights to control their own funds. That theory had two prongs: first, what the Terms of Service said about loans; and second, whether the "going negative" and "back door" features of the info@ account independently constituted theft by allowing Alameda to withdraw more than it had deposited.

At every turn, Judge Kaplan ensured Sam could not defend himself on these questions. He barred the defense from calling an expert on UK law—which governed the International exchange—who would have testified that the Terms of Service authorized the loans. He barred evidence that FTX's own lawyers had approved the loan arrangements. He barred testimony from the FTX attorneys who drafted the Terms of Service. The net effect was that the jury heard one side of the most important legal question at trial.

The fiat@ Account: Source of All the Trouble

Counter: the government argued this was not merely an accounting bug narrative; once leadership understood the shortfall, continued customer assurances and continued Alameda access could be seen as concealment of a real hole, not innocent operational delay.

The government repeatedly conflated two completely different arrangements: borrowing through the fiat@ account and borrowing through Alameda's trading account (the "info@" account). These involved different accounts, governed by different contractual arrangements, with very different legal consequences. The conflation was not accidental.

When FTX was founded, it could not get its own US bank accounts—American banks were reluctant to deal with international crypto companies. To solve this, FTX and Alameda entered into a payment agent arrangement: US customers deposited money into a bank account held by Alameda's subsidiary (North Dimension), and Alameda tracked the deposits in a separate "fiat@" ledger. FTX credited and debited customer accounts accordingly.

Under the payment agent arrangement, the balance Alameda owed FTX should have hovered near zero. That is what happened for the first year and a half. But a bug in the payment software caused US customer deposits to be posted to the fiat@ ledger as though they were inter-company transfers rather than pass-through deposits. The result: Alameda's stated balance grew from near zero to $4 billion by summer 2021, and from $4 billion to $9 billion by mid-2022.

Wang, Singh, and Ellison were each aware of this growing balance. None of them told Sam. Sam first learned of its true scale at the June 16, 2022 leadership meeting. He immediately took two steps: he put a team in charge of a project to reduce the fiat@ balance to zero, and he again urged Ellison to hedge Alameda's crypto holdings against further market crashes—which she had still not done despite two devastating crashes earlier that year. Both projects were underway when FTX filed for bankruptcy five months later.

B. Misrepresentation

Counter: the government's position is that reassurance statements mattered because they were made during a liquidity emergency to prevent withdrawals, and that a speaker with access to internal data cannot avoid fraud liability by claiming optimism when the omitted risk facts were mission-critical.

In addition to the misappropriation theory, the prosecution introduced a second theory mid-trial: that Sam misled customers about the risks the loans to Alameda imposed on them, in various public statements and in the Terms of Service itself. The prosecution assembled a long list of alleged misrepresentations—many too vague to constitute fraud, others true when made.

The most damaging, in the prosecution's telling, was a tweet sent on November 7, 2022—two days into the bank run:

The prosecution also argued that Sam misled customers in various blog posts and press statements about FTX's practices and protections. When the defense tried to introduce a contemporaneous blog post showing Sam's public position on FTX's ability to cover customer withdrawals, Judge Kaplan sustained the prosecution's objection and excluded it. The jury heard the government's characterization of Sam's beliefs; it was not allowed to see the primary evidence of what those beliefs actually were.

C. Evidence of Criminal Intent

Counter: intent can be inferred circumstantially from repeated high-risk choices: privileged code access for Alameda, borrowing that exceeded ordinary limits, selective public messaging, and emergency fundraising while publicly saying assets were safe. Under this view, motive need not be personal cash extraction; preserving empire and influence can itself supply motive.

The government's theory of intent was that Sam intended to defraud customers from the moment he and the co-founders structured the loans to Alameda, and that he knew the loans would ultimately harm customers. The defense's theory was that Sam believed throughout that Alameda would be able to repay its debts to FTX, and that FTX's collapse was the result of a bank run that might have been survived with adequate liquidity.

On motive, the government's case was remarkably weak. At the time of the alleged fraud, Sam and his co-founders were billionaires many times over. He could have sold five percent of his stake in FTX for over $1 billion if he needed personal funds. He had no financial motive to steal from his own customers—the theft, as the prosecution framed it, would have destroyed the very company that made him wealthy.

Wang acknowledged on cross-examination that at the time of FTX's collapse, Alameda was both solvent and fully liquid, with a surplus of about $250 million. The crypto market was at a cyclical low. When crypto assets recovered after the bankruptcy—as they have since done dramatically—the value of the estate recovered with them. The assets were never "gone." They were temporarily illiquid in the midst of a bank run. This is why 98% of creditors have been repaid in full, with interest.

The government's claim that FTX customers had suffered billions of dollars in losses—made in opening, in summation, and at sentencing—was devastatingly effective and factually wrong. The prosecution told the jury Sam had stolen and squandered customer money. The bankruptcy estate simultaneously made customers whole. The two facts were never reconciled for the jury, because Kaplan prohibited the defense from introducing evidence about the bankruptcy recovery.