A federal judge in New York dismissed fraud claims against Uniswap for the second time this month, and the decision carries implications far beyond the cryptocurrency industry.
At stake: whether platforms that provide neutral infrastructure can be held liable when bad actors exploit those tools to commit fraud.
Judge Katherine Polk Failla’s ruling applies a principle that translates cleanly across technology sectors: you don’t sue the New York Stock Exchange for selling you fraudulent stock.
The same logic, she argues, applies to decentralized exchange protocols.
However, as scams proliferate across digital platforms, courts are being forced to decide who should serve as the de facto insurer for internet-scale fraud. The FBI reported over $6.5 billion in losses from cryptocurrency investment fraud in 2024 alone.
The theory plaintiffs keep testing
The case began when investors who lost money on tokens traded through Uniswap’s interface attempted to shift liability from the scammers who issued worthless assets to the developers who built the trading rails.
Their legal strategy: frame the provision of market infrastructure as “aiding and abetting” fraud.
Failla rejected this approach in August 2023, writing that plaintiffs “are looking for a scapegoat” because “the defendants they truly seek are unidentifiable.”
The Second Circuit affirmed dismissal of federal securities claims in February 2025, stating it “defies logic” to hold smart contract developers liable for “a third-party user’s misuse of the platform.”
Undeterred, plaintiffs filed a second amended complaint in May 2025, pivoting to state-law theories.
They alleged that “in excess of 98%” of tokens traded through the interface were scams and claimed Uniswap collected over $100 million in fees from fraudulent activity.
This month, Failla also dismissed those claims, reportedly with prejudice. This means that the appeal clock now starts on what could become a controlling precedent.
Drawing the liability boundary
The legal principle at issue predates cryptocurrency by decades.
Courts evaluating secondary liability for fraud have consistently required two elements: specific knowledge of the wrongdoing and substantial assistance that materially aided the fraud.
Providing general-purpose infrastructure that scammers also happen to use doesn’t meet that standard.
The Supreme Court applied similar reasoning in Twitter v. Taamneh, rejecting attempts to hold social media platforms liable for terrorism merely because terrorists used their services.
The question in both contexts: does operating neutral infrastructure that enables both legitimate and illegitimate activity constitute meaningful assistance to wrongdoing, or does it simply make you the most convenient defendant with money?
Failla’s opinion confronts this directly. She notes that if anonymity in financial markets is “troublesome enough to merit regulation,” that decision belongs to Congress, not tort litigation.
The judiciary draws lines based on existing law; legislatures write new rules when policy demands change.
Why the stakes extend beyond DeFi
The “make the toolmaker pay” theory surfaces across technology litigation with striking regularity.
App stores face lawsuits over scam applications that slip through review processes. AI companies face liability demands when someone uses a language model to generate phishing emails. Payment processors defend against claims that they enabled fraud by processing transactions.
In each case, plaintiffs confronting uncollectable judgments against actual wrongdoers seek to recharacterize platform operators as perpetrators. The economic logic is straightforward: scammers vanish or have no assets; platforms have balance sheets.
However, treating infrastructure providers as insurers creates its own distortions.
Chainalysis estimates that crypto scams and fraud reached $17 billion in 2025. If courts assigned that liability to access layers rather than to perpetrators, platforms would face a binary choice: price insurance premiums into fees or gate access so aggressively that only pre-vetted activity occurs.
The fee uplift math is unforgiving. Monthly scam losses divided by legitimate volume, plus legal overhead and margin, compound quickly.
In fraud-intensive environments, even low single-digit liability exposure translates to material cost increases or hard curation, exactly the friction decentralized systems were built to eliminate.
The curation problem platforms face next
Even if neutral tools maintain liability protection, curated surfaces present different questions.
Featured token lists, promoted trading pairs, default routing algorithms, and “recommended” swap interfaces all involve editorial judgment.
Plaintiffs will argue that curation implies both knowledge and assistance, the two elements courts require for secondary liability.
This creates pressure for interfaces to either strip curation entirely or add compliance infrastructure. Token allowlists and denylists, pre-trade risk warnings, geographic gating, and enhanced due diligence all carry costs.
Some platforms may determine that operating as genuinely neutral rails, with no recommendations, no featured content, and no algorithmic optimization, provides the cleanest liability posture.
That defensive retreat has consequences. Users benefit from curation when it surfaces quality over noise. Markets function better with reputation signals and quality filters.
Yet, if providing those features converts a platform from neutral infrastructure to an active participant, rational actors will eliminate them.
Feature / behaviorNeutral infrastructure or curated?Knowledge signalAssistance signalWhy plaintiffs target itLikely defense framingUncurated swap interface / generic routingNeutralLowLowDeep-pocket “rails” defendant; argues access = facilitationGeneral-purpose tool used for lawful + unlawful activity; no specific knowledge; no material assistancePublic warnings / terms-of-service disclosuresNeutralLowLowTries to argue warnings were inadequate or misleadingDisclosures defeat deception/omission theories; information not unique/nonpublic; users assumed riskFeatured token listsCuratedMed–HighMed“You highlighted it” → implied endorsement; curation as participationUI sorting ≠ guarantees; no specific knowledge of fraud; standard informational displayPromoted pairs / paid placementsCuratedHighHighClosest to “substantial assistance” + motive; looks like sponsorshipClear labeling + separation of ads vs listings; no involvement in issuer misreps; compliance controls mitigate“Recommended” swapsCuratedMed–HighMed–HighRecommendation suggests suitability/endorsement; reliance + causation angleRecommendations are algorithmic UX defaults, not advice; disclaimers; no knowledge of specific schemeDefault routing algorithm optimizationsGray zone (lean curated)MedMedPlaintiffs claim routing “steered” them to scam liquidityRouting optimizes execution (price/liq), not token quality; content-neutral; no issuer coordinationAllow/deny lists (token gating)Compliance-heavy (both)MedLow–MedIf you can block, plaintiffs argue you had control/notice dutiesRisk controls reduce harm; lists are prudential safety measures; absence of listing ≠ endorsement; still no specific fraud knowledgeManual token review / “verified” badges (if applicable)CuratedHighHigh“Verification” implies diligence + relianceVerification scope is narrow (e.g., contract match), not investment quality; explicit criteria + disclaimersCustomer support escalation / internal reports handlingNeutral (process)Med–High (post-notice)Low–MedPlaintiffs argue notice = knowledge; failure to act = assistanceTiming matters: notice often after losses; no conscious avoidance; reasonable response policiesFee design tied to specific pairs/tokens (if applicable)Gray zoneMedMedArgues profit motive from fraud + incentive to keep listingsFees are transaction-based and content-neutral; no special relationship with issuers; not tied to misrepresentations
What courts are and aren’t deciding
Failla’s rulings don’t establish that platforms can indefinitely ignore fraud.
They establish that generalized awareness of bad actors using a system, rather than specific knowledge of particular scams as they occur.
They distinguish between operating lawful infrastructure that scammers also access and materially assisting specific fraudulent schemes.
The distinction matters because it preserves the ability to build general-purpose tools without underwriting every possible misuse. Hammers get used in construction and break-ins, and courts don’t assign liability to hardware stores.
The question is whether digital infrastructure deserves the same treatment or whether internet-scale fraud creates policy problems that require internet-scale solutions.
Plaintiffs’ lawyers will almost certainly appeal. If the Second Circuit affirms, the precedent hardens. Interface developers, wallet providers, and middleware infrastructure gain a clearer safe harbor.
Investment flows toward permissionless systems with reduced tail risk.
If the Circuit reverses or if legislators decide victims need solvent defendants regardless of what tort law says, the compliance burden shifts. Platforms adopt know-your-transaction regimes. Costs rise. Innovation migrates to jurisdictions with more predictable rules.
Who decides what happens next
The immediate procedural reality is that federal civil appeals must generally be filed within 30 days of the entry of judgment.
That creates a near-term catalyst for whether this becomes binding law or returns for another round of litigation.
The larger policy question extends beyond any single case. Failla explicitly flagged this in her original opinion: if lawmakers want different rules about anonymity and platform liability in financial markets, that’s a legislative decision.
Courts apply existing standards, while Congress writes new ones.
The current standard, knowledge plus substantial assistance, sets a high bar for plaintiffs seeking to relabel infrastructure as a perpetrator. It protects toolmakers who build neutral systems that enable both legitimate commerce and fraud. It forces victims to pursue actual wrongdoers rather than convenient corporate defendants.
Whether that standard remains adequate as scams industrialize and professionalize is the question Failla declined to answer.
Federal judges interpret the law as written. If the law should change because fraud has scaled beyond what existing liability frameworks anticipated, that’s a call for elected officials who write statutes, not appointed judges who apply them.
The decision matters because it determines who bears internet-scale fraud losses in an era when those losses are measured in billions annually.
Scammers vanish. Victims demand recovery. Platforms provide the most visible target. Courts now repeatedly say that visibility doesn’t equal liability, but the economic pressure to find someone who pays doesn’t disappear just because judges draw clear lines.
