When Short-Seller Targets Fight Back: Two SDNY Complaints Against Culper Research Over the AppLovin "Red Curtain" Report

On 10 June 2026 and 12 June 2026, two complaints were filed in the U.S. District Court for the Southern District of New York against Shadyside Partners LLC (d/b/a Culper Research) and its founder Christian Lamarco. The first, JC International Finance Limited et al. v. Lamarco (Case No. 1:26-cv-04930-JMF), brings 15 corporate plaintiffs across Hong Kong, the British Virgin Islands, the Philippines, Japan, and the Cayman Islands together with 2 individual plaintiffs in libel per se. The second, Tang v. Shadyside Partners (Case No. 1:26-cv-05018), is a single-plaintiff defamation per se action by Hao Tang, the Cypriot-citizen, Hong Kong-resident founder of the Goldenway Group. Both target the same Culper publication: a 30-page short report on AppLovin Corporation (NASDAQ: APP), titled "Behind the Red Curtain: CCP Intelligence, Human Trafficking, Money Laundering; Undisclosed Stock Pledges; Secret Chinese e-Commerce Deals," published on Culper's website and X account on 12 June 2025.

The complaints arrive 8 days after a federal jury in Manhattan convicted activist short seller Andrew Left of securities fraud on 2 June 2026. Both filings cite the conviction in their first substantive paragraphs. This post reads each complaint on its own terms, against the report and the underlying source documents it cites, and frames what the two filings, taken together, imply about the discipline of writing a short-seller report.

Short-seller reports are not neutral documents, and these complaints are not either. Culper discloses a short position in AppLovin and stood to profit if the share price fell. The plaintiffs are entities and individuals named in (or, on their reading, indirectly tarred by) the report, and they are seeking damages and injunctive relief. Both sides are litigants.

What we can do is take the report's specifically cited primary materials (the 2017 FCA Final Notice, the 2024 UK Employment Tribunal judgment, AppLovin's S-1 capital structure, the public corporate registries in Hong Kong and the Philippines) and read each complaint's specific paragraphs against those materials. We treat the complaints as filed allegations rather than findings of fact. The deeper question, which the cases raise even before either gets to a motion to dismiss, is what separates a defensible short-seller report from one that crosses the line into actionable defamation. The two complaints put that question on the docket in unusually concrete form.

Overview

1. The June 2025 "Red Curtain" report and the two June 2026 complaints
2. Tang v. Shadyside Partners: defaming a private investor by name
3. JC International Finance et al. v. Lamarco: the libelous-graphic theory
4. Four failure modes the complaints make legible
5. The Andrew Left conviction and Culper's own prior litigation record
6. The peer comparator: CapitalWatch retracted
7. The discipline of writing a short report
8. Where we sit

The June 2025 "Red Curtain" report and the two June 2026 complaints

The Culper report runs 30 pages on culperresearch.com and on the firm's X account. Its first-line disclosure states that the firm is short AppLovin and stands to profit from any decline in the share price; both complaints reproduce the disclosure. The report's headline contention is that AppLovin has been "backed since at least 2017 by Chinese national Hao Tang" and that the company "obfuscated disclosures" to hide that backing from investors.

Around that anchor, the report stacks three structural layers. The first is a catalogue of 14 factual statements about Tang personally, which the Tang complaint enumerates in its causes of action. The second is a visual "summary" on page 13 that arranges 15 named corporate entities and individuals into labelled boxes, with arrows leading to a single large red box of "Bad Acts"; the JC International Finance complaint is built around that page. The third is 30 pages of supporting research drawn from FCA filings, an English Employment Tribunal judgment, press reporting on Myanmar's KK Industrial Park, Philippine casino reporting, and FAA aircraft-ownership records.

Neither complaint is a securities-fraud action against Culper. Both are defamation actions filed by people and entities named in the report. The Tang complaint asserts a single cause of action for defamation per se under New York law, anchored in 14 specific statements. The JC International Finance complaint asserts a single cause of action for libel per se under New York law, anchored in the visual it calls the "Libelous Graphic." Different theories, same publication, same defendants, three calendar days apart.

The timing is, on the public record, structural. The Tang complaint was filed on the one-year anniversary of the report (12 June 2026), which matters for New York's one-year statute of limitations on defamation under CPLR § 215. The JC International Finance complaint was filed two days earlier, on 10 June 2026, comfortably inside the same window. Both filings are paced against the limitations period rather than against any new development in 2026.

Reading the two complaints together lets us see the report's structure at two levels: how it treats a named private individual at the centre of its narrative, and how it treats the broader cast of corporate entities arranged around him in a one-page visual.

Tang v. Shadyside Partners: defaming a private investor by name

Hao Tang is, on the publicly available record, a Cypriot citizen resident in Hong Kong who founded the Goldenway Group and, in June 2024, Tier One Investment Management, a Cayman-Islands-Monetary-Authority-licensed investment manager of which he is a director. He has, on his complaint's account, pursued a passive-investment philosophy for over a decade. AppLovin is one of his portfolio holdings.

Public records, the complaint pleads, show that he has never held Class B shares of AppLovin; at the IPO in April 2021, Class B shares controlled 93.4% of AppLovin's voting power. Tang has never been a director, officer, or employee of AppLovin. The complaint pleads that he has never met AppLovin's CEO and co-founder Adam Foroughi, and that he has not participated in the company's day-to-day operations or communications.

The Culper report places Tang at the load-bearing centre of its AppLovin thesis. It states that he "backed" the company and that AppLovin's disclosures around his ownership "suggest a cover-up." The Tang complaint catalogues 14 specific statements that, on its reading, accuse Tang personally of criminal conduct. The 14 statements fall into three buckets.

The first bucket is money laundering and fraud. The report writes that Culper's research "reveals Hao Tang's numerous ties to ... money laundering" and that "Tang's network is involved in money laundering" (Report at 2, 4, 29). It anchors these statements on two distinct underlying documents: the 2017 FCA Final Notice involving Goldenway's UK arm, and the 2024 judgment of the UK Employment Tribunal in Bhagani v. Goldenway Global Investments (UK) Ltd.

The Tang complaint's response (¶33 to ¶38) is mechanical. The 2017 FCA Final Notice denied a specific person's application to act as a compliance officer (Gregory Nathan) on competence grounds, and was issued five years before the whistleblower disclosures at Goldenway UK were even made. The 2024 Tribunal ruled only on whether the whistleblower's disclosures qualified as protected public-interest disclosures for the purpose of finding that his dismissal was improper. The Tribunal did not, the complaint pleads, make any finding that the underlying misconduct alleged in the disclosures had actually occurred.

The report frames the combined material as "the UK employment tribunal ruled in the whistleblower's favor, validating the claims." That sentence is the load-bearing one, and the complaint argues it substitutes one proceeding's name for another's findings. The matter sits at the centre of the complaint because it is, on the face of the cited documents, checkable.

The second bucket is human trafficking and murder-for-hire. The report writes that "Hao Tang's numerous connections tie to ... human trafficking ..., murder for hire" and that Tang is a "bad actor with extensive direct and indirect ties to ... human trafficking" (Report at 2, 4, 13). The complaint walks through the inferential chain at ¶41 to ¶47.

Goldenway Finance (rebranded JC International Finance after a third-party sale, which Tang's complaint pleads happened before the relevant loan was made) lent money to Huang Youlong. Youlong's cousin is She Zhijiang. She Zhijiang allegedly co-ran Myanmar's KK Industrial Park, which press reporting has called a "fraud factory" and a "human trafficking hub." KK is associated, in press reporting, with former Chinese triad leader "Broken Tooth." The Tang complaint reads the chain as four removes from Tang himself.

The murder-for-hire claim adds another step on top of that chain. Liduan Wang, who served at one time as chairman of FortuneGate Holdings Philippines (a casino-management company that one Goldenway announcement once erroneously identified as part of the Goldenway Group), was reportedly tied to the Nine Dynasty Casino junket, which press reports linked to the murder of Chinese steel magnate Anson Que. The complaint's reading is that none of these steps establishes Tang's personal involvement in or knowledge of human trafficking or murder; each step is at most a commercial counterparty relationship one or two removes away from a press allegation about a third party.

The third bucket is CCP intelligence and propaganda. The report writes that "Hao Tang holds numerous ties to Chinese intelligence [and] CCP propaganda outlets" and that Tang "appears to be linked to the 2023 disappearance of former CCP state journalist Fu Xiaotian and State Councilor Qin Gang, suggesting he retains high-level CCP connections" (Report at 2, 4, 16). The asserted basis is FAA records that, on the report's reading, show a Goldenway-operated jet was the aircraft Fu was last seen boarding before her disappearance.

The complaint's response (¶51 to ¶52) is that operating a registered jet on which a third party flew, without anything tying Tang personally to the flight's authorisation or purpose, is not evidence of intelligence ties. Tang pleads that he does not know either Fu or Qin beyond what is in the public record. The argument is that aircraft ownership and intelligence work are different categories of claim.

The Tang complaint's load-bearing legal argument is structural rather than evidentiary. It does not contest most of the underlying counterparty relationships that Culper documents. It concedes, in detail, that some of Tang's corporate vehicles had commercial dealings with the people the report names.

The complaint's argument is that the report's claim is categorical (Tang is a "bad actor," involved in money laundering, trafficking, and intelligence work) while the cited evidence is relational (Tang owned a vehicle that had a transaction with someone whose cousin is alleged to be a criminal). On New York's defamation-per-se standard, statements that impute serious criminal conduct, or that injure a person in their trade or business, are pleaded without special damages. The complaint pleads both prongs.

JC International Finance et al. v. Lamarco: the libelous-graphic theory

The JC International Finance complaint is a different shape. It collects 17 plaintiffs, 15 of which are corporate entities incorporated in Hong Kong, the British Virgin Islands, the Philippines, Japan, and the Cayman Islands. Its target is not the text of the Culper report. It is a single page in the report: page 13, on which Culper places a visual that the complaint calls the "Libelous Graphic."

The graphic, as the complaint describes it (¶3 to ¶6, ¶28 to ¶29), arranges 15 entities and individuals into separate labelled boxes. Arrows from each labelled box converge on a single large rectangle at the bottom of the page. That rectangle, set in red type, lists a single block of "Bad Acts": "Money Laundering," "Forex Fraud," "Disappearances," "Human Trafficking," "POGO Raids" (a reference to Philippine offshore gaming operations), and "14K Triad."

The Bad Acts are not subdivided. They are not connected with named-entity-specific arrows. Every plaintiff's box ultimately points, directly or transitively, to the same red box containing every Bad Act.

The complaint's pleading on this point invokes what New York courts have called the "headline rule" (¶8): a defamatory summary or headline can be actionable independently of the body of the article, on the theory that many readers do not read past the headline or the visual summary. The body of the Culper report, the complaint argues, makes clear that most of the listed plaintiffs have no described connection to most of the listed Bad Acts.

The pleading walks through specific examples. The body's discussion of Discovery Key, described as one of Tang's offshore entities, makes no connection between Discovery Key and any of the Bad Acts (¶31). The body's discussion of Southwind Media is limited to its acquisition of MediaPro and a content agreement with China Media Group; no Bad Acts are alleged in the body (¶32). The body's only link between JC International Finance and "human trafficking" is the alleged loan to Huang Youlong, whose cousin is connected by press speculation to KK Industrial Park (¶34); no link in the body connects JC International Finance to "Money Laundering" or "Forex Fraud" (¶36). The body's only link between IEC and any Bad Act, the complaint pleads, is non-existent (¶37). The body's only link between Ho Wong Meng (FortuneGate's CEO) and any Bad Act is that Liduan Wang served at one time as chairman of the same entity (¶38).

The libelous-graphic theory is structurally interesting because it severs the visual layer from the textual layer of a research report. The visual is treated as an independent publication. If the complaint's reading of the graphic is right (the panel reproduces the graphic at ¶29), then the report contains a single visual on a single page that, on its face, attributes every listed criminal act to every listed entity, while the body of the report does not.

Courts applying the headline rule typically require a showing that the visual or headline says something the body does not, and that the false impression created by the visual is not corrected by the body in any meaningful sense. The complaint pleads both prongs. Whether the court agrees that the graphic, as a matter of New York law, says what the complaint says it says is a question on which a motion to dismiss may turn.

The plaintiff list itself is a useful diagnostic. Some plaintiffs (JC International Finance, Southwind Media, FortuneGate Holdings Philippines, the Goldenway entities) are named in the body of the report with at least some specific allegation attached. Others (Discovery Key, Prominence Trust, Angel Pride, Midterm Success, Asian Creative Capital, IEC, Fenton Capital, and the two individual plaintiffs) appear in the graphic but, on the complaint's account, do not have any specific Bad Act tied to them in the body. The complaint asks for damages of "no less than $1,000,000," permanent injunctive relief, and other relief.

Four failure modes the complaints make legible

Whatever one thinks of either side's chances at trial, the two complaints together surface four specific failure modes that are useful to name. Each one is concrete enough to be checked in a draft.

The first is conflating two separate proceedings into one "validated" claim. The Tang complaint's clearest example (¶33 to ¶38) is the report's combination of the 2017 FCA Final Notice (denying Gregory Nathan's application to act as a compliance officer at Goldenway UK on competence grounds) with the 2024 UK Employment Tribunal judgment (finding that a Goldenway UK employee's whistleblower disclosures were qualifying and that his dismissal was improper). The report frames this as "the UK employment tribunal ruled in the whistleblower's favor, validating the claims." The complaint's reading is that the Tribunal validated the claims as protected disclosures, not the underlying allegations as factual.

If true, the conflation collapses two proceedings, on different facts, in different fora, addressing different legal questions, into one wrong sentence. The fix is reading what the Tribunal judgment actually held, which is published, and writing a sentence that matches its actual holding.

The second failure mode is building a guilt-by-association chain that the reader has to count to follow. Tang lent (via a corporate vehicle, after a corporate sale) to A; A's cousin B is alleged to run C; C is associated with D; D is reportedly tied to E; E is linked to a murder. Two of those links are press reports about third parties. None of them establish anything about Tang personally.

The further you have to walk down such a chain to get to the named subject of a report, the harder it is to defend any categorical conclusion at the end of the walk. Both complaints lean on this structural critique because the cited material does not, on the plaintiffs' reading, ever close the gap between the chain and the named subject.

The third failure mode is the summary visual that the body of the report does not support. The Libelous Graphic on page 13 is the heart of the JC International Finance theory: every plaintiff in a separate labelled box, every Bad Act in a shared, undifferentiated rectangle, every arrow ultimately landing in the rectangle. If the body does not establish a particular plaintiff's link to a particular Bad Act, the graphic does work the body declines to do.

Headline-rule case law, in New York and other jurisdictions, exists precisely because readers retain summary visuals and headlines at much higher rates than they retain qualified textual claims. A defensible visual summary, on this reading, would subdivide the Bad Acts by named entity, or omit entities for whom the body offers no link.

The fourth failure mode is attacking a private investor by name with categorical criminal allegations. AppLovin is a public company; its disclosures are public material. Tang is, on his pleading, a private citizen, not an officer or director, who has never met the CEO. New York defamation law treats public figures, limited-purpose public figures, and private figures differently for the purpose of the actual-malice standard.

Private figures need to plead negligence as to falsity rather than actual malice, which means the standard of care the plaintiff has to clear is lower. A short-seller report that makes a private investor the centre of its public-company thesis is making a strategic choice about who to name. The Tang complaint argues that the choice carries with it the burden of a private-figure defamation standard.

The Andrew Left conviction and Culper's own prior litigation record

The litigation environment in which both complaints land has changed in the last year. On 2 June 2026, a federal jury in Manhattan convicted activist short seller Andrew Left of securities fraud. Both complaints cite the case by name in their first substantive paragraphs (Tang Complaint ¶22 n.2; JC International Finance Complaint at p.2 n.2).

The Left case turned, on the publicly available record, on the alleged inconsistency between his published recommendations and his actual trading positions. The substantive standards in defamation per se and securities fraud are different. The procedural and reputational pressure on short-seller research methodology, in light of a successful federal-jury conviction of a high-profile practitioner of the genre, is not.

The JC International Finance complaint also documents Culper's own prior litigation record (n.3): LifeMD, Inc. v. Lamarco (W.D. Pa. 2021, subsequently dismissed), Gorilla Technology Group, Inc. v. Shadyside Partners (S.D.N.Y. 2025, subsequently settled), and Tecnoglass Inc. v. Lamarco (S.D.N.Y. Sept. 2025, voluntarily dismissed). The complaint quotes a 20 March 2024 Culper X post comparing the firm to the Grim Reaper going door to door (Tang Complaint ¶24).

Neither complaint argues that the prior actions were resolved against Culper on the merits. The pattern matters as colour rather than as decided cases. Defamation defendants whose names recur on dockets attract a different reading from courts than first-time defendants do, and the recurrence is now in the public record.

The peer comparator: CapitalWatch retracted

The Tang complaint pleads a comparator that is harder to wave off than the docket history. After publication, Tang sent letters identifying alleged inaccuracies to Culper and to another short seller, CapitalWatch, which had published "some of the same false statements" about Tang (¶61). CapitalWatch retracted.

Its retraction stated that "descriptions asserting direct connections between Mr. Tang and [various individuals] were inaccurate and failed to meet our publication standards." CapitalWatch also apologised for "the distress caused and the potential impact on his personal reputation." CNBC, Yahoo Finance, and Investopedia reported on the retraction.

Culper, on the complaint's pleading, did not retract. Tang's lawyers sent three demand letters: in December 2025, on 8 January 2026, and on 22 January 2026 (¶55 to ¶60). Culper responded only to the third and declined to retract or correct any statement.

The peer comparator matters because it shows that a contemporaneous reader at another short-seller publication, presented with the same allegations and the same demand letter, reached a different editorial conclusion. It is weaker than it first looks, though. Actual malice turns on what Culper itself knew or recklessly disregarded, not on what a peer decided, and CapitalWatch's caution could reflect a lower risk appetite, a thinner underlying file, or simple prudence rather than a judgment that the claims were false. A peer retraction is suggestive on the malice question, not dispositive. Culper will argue it shows only that two publishers priced the same litigation risk differently.

The discipline of writing a short report

Stepping back from the two specific cases, the discipline a short report asks its author to hold is unusual. Short reports do work few other published documents do. They aggregate publicly available filings, primary documents, source interviews, and inference into a thesis that is meant to move the share price of a listed company.

The author has a directional financial interest. The standard of care is therefore higher than for a news article on the same facts, not lower. The First Amendment and the New York Times Co. v. Sullivan line of cases create breathing room for vigorous public-interest journalism. That room shrinks when the subject of the report is a private individual, when the language used is categorical rather than scoped, and when the publisher's economic interest is direct rather than incidental.

Culper has real arguments on the other side, and a motion to dismiss is where many defamation claims of this kind end. Several of the report's statements may be defensible as opinion or rhetorical hyperbole rather than provable fact, and the counterparty relationships it documents may be substantially true even where the inferences drawn from them are contested. Under New York law, substantial truth is a complete defense, and a statement that cannot be proven false is not actionable. Culper will also invoke the public-interest privilege and the Sullivan line to protect aggressive reporting on who owns and influences a large public company, and it will argue that an investor whose stake and alleged sway over that company are themselves matters of public controversy is at least a limited-purpose public figure, which would force Tang to plead actual malice rather than negligence. Whether Tang is a private figure or a limited-purpose public figure is genuinely contested, and the answer shapes the whole case.

None of that makes the complaints frivolous, and none of it makes Culper's defense a sure thing. The point is narrower: the line between a defensible short report and an actionable one runs through several unsettled questions of New York law. A short report on the same broad set of facts the Culper "Red Curtain" report engages with could, on the discipline the two complaints implicitly call for, lower its litigation risk in three ways.

First, claims would be scoped to what the cited evidence supports. "Goldenway UK was the subject of a 2017 FCA Notice denying a specific officer's application on competence grounds, and a 2024 UK Employment Tribunal found a whistleblower's disclosures qualifying, the underlying allegations of which have not been adjudicated" is a defensible sentence. "The UK tribunal ruled in the whistleblower's favor, validating the claims" is, on the Tang complaint's reading, not.

Second, guilt-by-association chains would be presented as such, not collapsed into categorical attributions. A loan from a Tang-affiliated lender to a person whose cousin is alleged to run KK Industrial Park is, on its own, an attenuated chain. The report's job, on a defensible standard, is to say so, and to scope the inference to what the chain can carry.

Third, summary visuals would correspond, item by item, to the body of the report. Entities for whom no specific Bad Act is established in the body would not appear in the same row as entities for whom one is. The headline rule exists because the visual layer of a report often does the most reader-facing work, and is therefore the layer where overreach is most costly.

None of those constraints would have prevented the Culper report from making its underlying case for shorting AppLovin. The report's core economic thesis (that AppLovin has undisclosed Chinese-investor exposure, that this is material to its TikTok-related acquisition strategy, and that the public disclosure record under-states the relationship) is independent of the Tang-as-criminal framing. A version of the report that ran the same financial-disclosure analysis without the personal allegations would be a less viral document, and would probably move the share price less on day one. It would also carry materially less litigation risk on day 365.

Where we sit

Short disclosures and short campaigns are available on Plainview, our risk research platform for investors. Culper runs autonomous agents that hunt risk across enterprise systems. Short reports are one input we test against the primary documents behind each claim.

The two Culper-AppLovin complaints are useful to anyone reading short reports professionally because they make the failure modes legible: a conflated tribunal citation, a guilt-by-association chain, a summary visual the body does not support, and a categorical attack on a private investor by name. None of these are new failure modes. The cases give them a docket number.

If your team is reading the Culper report (or any comparable single-name report on a U.S.-listed company with foreign-investor exposure) and would find it useful to compare notes on how to scope each underlying claim against its primary documents, write to hello@bollwerk.ai.