Sokolowski, Pt. 2
Cyborg David against Goliath (spoiler: Goliath probably wins)
I'm returning to my post from a few weeks ago about Steve Sokolowski and his brother Chris, who filed a pro se complaint in Pennsylvania that received mockery from certain corners of the legal internet. In my last post, I explained why the AI tool they were using (OpenAI's o1 model) represented a meaningful improvement over prior versions of ChatGPT. At the time I wrote, o1 showed promise, was still a curiosity and cost $200/month. As I went to publish the post, DeepSeek released its own o1 competitor, which is free, open-source, and works just as well. Since then, major American AI companies have cut their prices and raced to release their own "reasoning model"; in six weeks, o1 went from a curiosity to a yardstick.
This post is about the complaint itself, and whether I think it’s any good. But first, I want to discuss what the situation is like for the pro se, or unrepresented civil litigants who, unlike the Sokolowskis, are not grinding out 100-page complaints with premium OpenAI accounts.
The Current Situation for Unrepresented Litigants
Most (or many) people know that, thanks to a 1963 Supreme Court case, Gideon v. Wainwright, criminal defendants have a constitutional right to a lawyer. Civil (non-criminal) litigants have no such right. In 1981, in Lassiter v. Dep't of Social Services, the Court held that a parent in civil proceedings to terminate her parental rights did not have the right to an attorney. This left the door open to a constitutional right in certain "quasi-criminal" proceedings, which the Supreme Court effectively closed in Turner v. Rogers, a 2011 case where the Court held that a noncustodial parent did not have a constitutional right to an attorney at a civil contempt hearing for nonpayment of child support, even when the noncustodial parent faced the potential penalty of incarceration.
In the United States, most litigation is civil, and the vast, vast majority of litigation takes place in state courts that handle all kinds of disputes, from car wrecks to foreclosures to debt collection and everything in between. Federal courts, despite their outsized importance in the public imagination, have a more limited jurisdiction and docket. To illustrate the difference in volume: The number of cases filed annually in New York City housing court is larger than the number of civil cases filed across all U.S. federal courts.1
Because of variations in data quality across courts (a topic for another time), we do not know exactly how many state court cases involve unrepresented litigants. And because there are so many state courts—far more than 50, because states often have different types of courts for different types of cases—comprehensive studies are very difficult. The most recent comprehensive2 study was conducted by the National Center for State Courts in 2015, using data from 2013-15; The most recent survey before that was in 1992.
Nevertheless, the two surveys show that the share of unrepresented litigants has increased considerably. The 2015 survey show that about half of defendants in civil cases were unrepresented, as compared to the 1992 study, which found 97% were represented.
The change reflects, more than anything, profound shifts in the types of cases that are heard in state court. In 1992, automobile personal injury cases dominated the docket; because of state-level insurance requirements, defendants are usually represented in these cases, and the prospect of insurance company payouts makes these cases remunerative for plaintiffs' lawyers. Now, the most commonly heard cases are debt collection (where a represented creditor sues an often-unrepresented debtor), eviction and foreclosure (where a represented landlord/bank sues an often-unrepresented tenant/homeowner), and domestic disputes (including divorce but matters like orders of protection, where it's not uncommon for both parties to be unrepresented). These are cases where there is not obvious remuneration for a private lawyer unless the dollar amounts involved are large.
In federal courts—where the Sokolowski brothers filed their case—the data is more comprehensive and the situation is less dramatic. But the numbers are still significant: The analysis in this 2018 law review comment suggests that the number of nonprisoner cases involving at least one unrepresented party hovers around 10 percent, across all courts. Compared to state court, the plaintiff-defendant valence is also flipped: Most unrepresented litigants are the plaintiffs who bring the cases (usually for civil rights or employment issues).
On the whole, it is almost certainly the case in March 2025 that a randomly selected civil litigant in the United States will not have a lawyer. This is a problem, not least because the entire system is built around the assumption that lawyers will be navigating it. Among people who think about this problem, there appear to be three general categories of approaches to solving it:
Continuing the fight for "civil Gideon." Although it's unlikely, based on the cases mentioned above, that the Supreme Court will enshrine a literal "civil Gideon" right anytime soon, there are various state and local pushes for "civil right to counsel" in certain types of cases.
"Unbundled" representation. Acknowledging that civil right to counsel is likely infeasible, this is a sort of triage approach that tries to provide partial legal services to more people.
"Demand-side" reforms. I'm borrowing this term from a 2015 law review article by Jessica K. Steinberg, a law professor. She suggests that rather than "supply-side" solutions focused on increasing the number of lawyers, we should address the unrepresented litigant crisis by reforming courts themselves to make them more accessible to non-attorneys.
On a personal note, in my own experience clerking at a federal district court, I observed that managing the prisoner and social security dockets—where most unrepresented litigants are found—taxed court resources much more heavily than the "ordinary" run of cases. As a law clerk, my job was to analyze the legal arguments of the parties and recommend to the judge how to rule. You quickly appreciate good lawyer is more valuable than a bad lawyer, and a bad lawyer is (usually) better than no lawyer not because good lawyers always have the better case, but because good lawyers make the process more efficient.
AI to the rescue?
That brief overview provides a baseline for evaluating the quality of the Sokolowski brothers AI-drafted complaint. Assuming we will not get civil Gideon anytime soon, and assuming that regulations will not be radically adjusted to increase the supply of lawyers, we should be open to using any technology that leads to a better experience for unrepresented litigants. Accordingly, in reading this complaint, the baseline should not the one assumed by Above the Law: Is this complaint better than something a lawyer could write? Instead, we should assume that the Sokolowski brothers would have filed this complaint without a lawyer anyway, and ask: Does the use of AI help the court achieve a more fair and efficient resolution than it would otherwise?
This is not a blow-by-blow summary. Because the internet never fails us, you can read a Bluesky thread from an attorney who did that, more or less. Although we differ in our snark quotient, he identified all the same issues I did when I first read it, so our analysis is not far off.
What is the case about?
As mentioned last time, the Sokolowski brothers make their bread running a cryptocurrency mining pool out of their house in State College, Pennsylvania. In what seems—even without the benefit of hindsight—like an unwise financial decision they they placed over 90 percent of their net worth (including cryptocurrencies and US dollars) into an account managed by Genesis Global Capital, LLC, a cryptocurrency bank. To do this, the Sokolowskis formed a company, Cryptocurrency Management, LLC (CM LLC), to execute a single promissory note with Genesis.
The Sokolowskis allege consumer fraud on behalf of themselves (not their LLC), arguing that Digital Currency Group, Inc. ("DCG"), Genesis's parent company, misled them by using fabricated balance sheets to assure the Sokolowskis that the loans were stable. As the complaint puts it, the Sokolowskis believed that they were engaging in a transaction "akin to a Certificate of Deposit or interest-bearing financial service for their household assets."
Although the complaint presents a coherent narrative, my lawyer alarm bells went off because it was clear that i) this whole case is about a contract between CM LLC and Genesis, but ii) neither Cryptocurrency Holdings or Genesis is a party. The plaintiffs are the Sokolowski brothers, as individuals, and the defendants are DCG, and the CEOs of both Genesis and DCG.
The punchline is that Genesis filed for bankruptcy in January 2023, when the crypto market bottomed out after the collapse of FTX and Sam Bankman-Fried. As we covered in the Mata v. Avianca case, the purpose of bankruptcy is to consolidate all claims against an insolvent debtor into a single proceeding. To make this effective, any claim brought by a party outside the pending bankruptcy proceeding is automatically stayed and creditors are effectively forced to participate in the bankruptcy if they want resolution. At the very end of the Complaint, we learn that CM LLC (i.e., the Sokolowskis) sold its bankruptcy claim against Genesis to "Jefferies Leveraged Credit Products LLC" on January 27, 2023 to pay "approximately $300,000" owed to Wells Fargo, money that was "collateralized against their long-term stock portfolio, as a result of a June 2, 2022 home purchase."
Is this a problem?
Yes. I agree with my counterpart on Bluesky that this is probably fatal to the claim (and would explain, at least in part, why no lawyer was willing to represent them). To be clear, it is legal to sell the claim, but, understandably, the system does not provide an easy route for someone who sells a claim regarding a contract to go back and sue over the same contract. To work around this, the Sokolowskis try to introduce facts showing that DCG, as well as the individual CEOs of Genesis and DCG, participated personally in the effort to misrepresent the nature of Genesis's investment portfolio.
To prove this, the Sokolowski brothers quote extensively from documents in an enforcement action brought by the New York attorney general case against Genesis. In his original Twitter post, Steve Sokolowski noted that they used AI to summarize all documents from this case and the Genesis bankruptcy. And honestly, I found it quite effective. I am not convinced they will win, but the argument is respectable, and they gather quite a few admissions by the Genesis and DCG executives that show that DCG made efforts to cover up the fact that Genesis was horrendously undercapitalized.
Even if they prove this, the bigger problem for the brothers is that they formed an LLC to deposit the money. There is a legal doctrine known as "piercing the corporate veil" which allows for limited exceptions to the general rule that corporate shareholders (known as "members" when the entity is an LLC) cannot be held individually liable for the actions of corporations. The doctrine varies by state, but usually you have to show, in effect, that the use of separate corporate entities was a sham. In almost every case I'm aware of, this argument is advanced against a party. Here, the Sokolowskis try to argue that their own LLC was a sham. I truly have no idea whether this can, or would work, but needless to say I'm skeptical.
Because the claim was sold, I have a hard time seeing how the case will not be dismissed. Even if the court accepts both the individual liability arguments for the defendants and the reverse veil-piercing arguments for the plaintiffs (unlikely), the misrepresentation argument itself feels pretty thin. There isn't great evidence that the promissory note itself concealed the fact that these were high-risk investments and, if the case reached discovery, it seems like the Sokolowskis—as people who work in the crypto industry—would have a hard time disclaiming that they were misled about the risk of crypto assets.
Is the complaint any good?
Judged by the standard of an “average” complaint: Yes. There are definitely some weird parts. It has an overly long discussion of simple topics like venue3, turning what should be a boilerplate paragraph into a 3 page memo. And the counts (the part where you actually assert the legal claims you are making) are too expository. They read like arguments you would make in a brief, with extensive case citations. This is unnecessary and generally a bad idea, because every paragraph is now an admission that the Sokolowskis will have to walk back if they want to make a different argument down the road. Strategically, you want to write a complaint that discloses the least amount of information needed to make a claim and save your exposition for later.
But, reading this casually, I was able to easily follow the narrative and the argument they were making. I think it's a bad argument, but the complaint makes decent version of that bad argument.
Which speaks to a point about unrepresented litigants, generally: Many of them don't have winning claims. But for good reason, our courts do not condition access based on a prediction about whether a complaint will succeed. A system that provides fulsome representation to unrepresented litigants is not better if more litigants win their cases (necessarily); it will be better because it ensures that all litigants are playing by the same rules. And in such a system, courts themselves are more efficient and the people who work in courts happier; this complaint is much better than some I read as a law clerk, including some submitted by lawyers.
For many reasons, I don't think the federal courts should start handing out OpenAI accounts to unrepresented litigants tomorrow, but this complaint is a small, hopeful sign that AI tools may be a useful tool for helping unrepresented litigants in the long run.
This figure comes from the Jessica K. Steinberg article on demand-side reforms, linked below.
The studies are “comprehensive” in that they fully analyze data from a few, select courts. However, they do not study every court.
In federal court, you have to allege that there is a reason to bring the case in the specific federal court district, or “venue.” In this case, it’s the Middle District of Pennsylvania, where the Sokolowskis live (a good reason).