SEC v. JIAN WU, No. 1:25-cv-07573, Southern District of New York (Sept. 11, 2025) — Complaint

1
Lee A. Greenwood
Christopher M. Colorado
Brian A. Kudon
David F. Benson*
Attorneys for Plaintiff
SECURITIES AND EXCHANGE COMMISSION
New York Regional Office
100 Pearl Street, Suite 20-100
New York, NY  10004-2616
212-336-9143 (Colorado)
[email protected]
*Pro Hac Vice Application Forthcoming
UNITED STATES DISTRICT COURT
SOUTHERN DISTRICT OF NEW YORK
SECURITIES AND EXCHANGE COMMISSION,
Plaintiff,
-against-
JIAN WU,
Defendant.
COMPLAINT
1:25 Civ. 07573
JURY TRIAL DEMANDED
Plaintiff Securities and Exchange Commission (“Commission”), for its Complaint against
Defendant Jian Wu (“Wu”), alleges:
INTRODUCTION
1.From 2021 to 2023, Jian Wu, a former employee of Two Sigma Investments, LP
(“TSI”), orchestrated a scheme to deceive his employer and reap millions of dollars in ill-gotten
gains by manipulating computer-based algorithmic investment models (“Models”) that TSI and
its affiliate, Two Sigma Advisers, LP (“TSA,” and, together, “Two Sigma”), used to make
investment decisions for their clients, including decisions to buy and sell securities.

2

2. Two Sigma developed and relied on Models to generate predictions and trading
signals for securities prices, which it aggregated into “Consolidated Forecasts” that sought to
predict the future performance of specific securities.  Two Sigma then relied on Models and
Consolidated Forecasts to decide whether to buy and sell securities for its advisory clients.
3. Two Sigma required that each new Model it developed and approved be
sufficiently uncorrelated to, or “decorrelated” from, existing Models and Consolidated Forecasts
to ensure that new Models generated unique forecasts that complimented, rather than duplicated,
the firm’s existing predictions.    Two Sigma thus sought to increase the likelihood that new
Models generated “alpha”—investment returns relative to the return of a relevant market
benchmark—that was not already captured by its existing Models and Consolidated Forecasts.
4. In violation of Two Sigma’s policies, between November 2021 and August 2023
(the “Relevant Period”), Wu secretly manipulated at least fourteen Models he created or helped
create, making unauthorized and undisclosed changes to those Models to increase their
correlation to the firm’s other Models and Consolidated Forecasts.  Thus, instead of generating
the unique trading signals and forecasts that Two Sigma expected, Wu’s manipulated Models
effectively replicated the output of other Models and Consolidated Forecasts.
5. Wu misrepresented to Two Sigma that his Models fell below the firm’s applicable
correlation thresholds and were generating unique forecasts, deceiving Two Sigma into relying
more heavily on them.  Consequently, when it deployed Wu’s Models, Two Sigma bought and
sold securities for its clients in amounts, concentrations, and frequencies that differed from its
intended strategies.  Wu’s secret Model changes caused c ertain Two Sigma clients to trade
securities in a manner that caused harm to them of at least $165 million during the Relevant
Period, which Two Sigma subsequently repaid.

3

6. Wu’s scheme also caused Two Sigma to falsely believe that his Models were
generating substantially more alpha not captured by the firm’s other Models and Consolidated
Forecasts than his Models actually generated, which caused Two Sigma to pay Wu millions of
dollars in cash bonuses and performance grants during the Relevant Period that it would not
otherwise have paid him.
7. Wu’s unauthorized and undisclosed changes to his Models and misrepresentations
to Two Sigma about his Models’ correlations violated the antifraud provisions of the Securities
Act of 1933 (“Securities Act”) and the Securities Exchange Act of 1934 (“Exchange Act”).
NATURE OF THE PROCEEDINGS AND RELIEF SOUGHT
8. The Commission brings this action under the authority conferred upon it by
Securities Act Sections 20(b) and 20(d) [15 U.S.C. §§ 77t(b) and (d)] and Exchange Act Section
21(d) [15 U.S.C. §§ 78u(d)].
9. The Commission seeks a final judgment:  (a) permanently enjoining Wu from
violating the federal securities laws and rules this Complaint alleges he has violated; (b) ordering
Wu to disgorge all ill-gotten gains he received as a result of the violations alleged here and to pay
prejudgment interest thereon, under Exchange Act Sections 21(d)(3), 21(d)(5), and 21(d)(7)
[15 U.S.C. §§ 78u(d)(3), 78u(d)(5), and 78u(d)(7)]; (c) ordering Wu to pay civil money penalties
under Securities Act Section 20(d) [15 U.S.C. § 77t(d)] and Exchange Act Section 21(d)(3) [15
U.S.C. § 78u(d)(3)]; (d) permanently prohibiting Wu from, directly or indirectly, acting as, or
being associated, with any investment adviser, under Securities Act Section 20(e) [15 U.S.C.
§ 77t(e)] and Exchange Act Section 21(d)(2) [15 U.S.C. § 78u(d)(2)]; and (e) ordering any other
and further relief the Court may deem just and proper.

4

JURISDICTION AND VENUE
10. This Court has jurisdiction over this action under Securities Act Section 22(a)
[15 U.S.C. § 77v(a)] and Exchange Act Section 27 [15 U.S.C. § 78aa].
11. Wu, directly and indirectly, has made use of the means or instrumentalities of
interstate commerce or of the mails in connection with the transactions, acts, practices, and
courses of business alleged herein.
12. Venue lies in this District under Securities Act Section 22(a) [15 U.S.C. § 77v(a)]
and Exchange Act Section 27 [15 U.S.C. § 78aa].    Wu transacted business in this District during
the Relevant Period.  In addition, certain of the transactions, acts, practices, and courses of
business alleged in this Complaint occurred within this District, including because Two Sigma,
where Wu was employed and to which Wu made materially false and misleading statements, had
offices located in this District to which Wu was assigned.
DEFENDANT
13. Jian Wu, age 34, is a permanent legal resident of New York, NY, and a citizen of
China.  He holds Ph.D.’s in Industrial Engineering and Operations Research from the University
of Southern California and Cornell University, respectively.  Wu was employed by TSI from
April 2018 to August 2023, holding various positions, including quantitative researcher, vice
president, and senior vice president.
RELEVANT NON-PARTY ENTITIES
14. TSI is a Delaware limited partnership headquartered in New York, NY.  TSI was
founded in July 2001 and has been registered with the Commission as an investment adviser since
August 2009.  TSI provides advisory services on a discretionary basis to various clients, including

5

private investment funds.  According to its Form ADV dated July 25, 2025, TSI had regulatory
assets under management of more than $ 110 billion.
15. TSA is a Delaware limited partnership headquartered in New York, NY.  TSA was
founded in December 2001 and has been registered with the Commission as an investment adviser
since February 2010.  TSA provides advisory services on a discretionary basis to various clients,
including private investment funds, a registered investment company, foreign funds, and separately
managed accounts (“SMAs”).  According to its Form ADV dated July 25, 2025, TSA had regulatory
assets under management of more than $74 billion.
STATEMENT OF FACTS
I. Two Sigma’s Model Development and Approval Process
16. During the Relevant Period, TSI developed Models and then licensed those
Models to TSA.  Both TSI and TSA used Models and Consolidated Forecasts to make investment
decisions for their clients, including buying and selling equity securities listed on exchanges such
as the New York Stock Exchange.
17. Two Sigma required new Models and changes to existing Models to be
sufficiently distinct from its other Models and Consolidated Forecasts, such that new or modified
Models were likely to generate alpha not already captured by the firm’s other Models or
Consolidated Forecasts.
18. During the Relevant Period, Two Sigma’s process for developing and approving
new Models or Model changes was described in its “Productionalize a Model” (“PA M”) manual,
which set forth the steps Two Sigma required before approving a new Model or Model change
(referred to collectively as the “PAM process”).

6

19. As part of the PA M   process, Two Sigma required modelers to submit Model
proposals or proposed Model changes through an internal system called “TOM.”  These required
submissions included detailed information about the Model proposal, such as its expected
correlation to existing Models and Consolidated Forecasts based on simulations run on historical
securities trading data.
20. Two Sigma also required the lead modeler for each Model to certify that the
information submitted with the TOM proposal, including all Model inputs, was accurate and
complete.
21. Two Sigma management then reviewed the TOM submissions and generally
approved only new Models or Model changes that met specific criteria, including a sufficient
lack of correlation to the firm’s existing Models and Consolidated Forecasts.
22. Once a new Model or Model change was approved, Two Sigma’s portfolio
management teams determined how to incorporate the Model into the firm’s investment
strategies for their advisory clients.
23. By following the PAM process, Two Sigma aimed to maintain an uncorrelated set
of Models that would generate unique investment returns, thereby enhancing the overall
performance of its advisory clients’ investments.
II. Two Sigma’s Storage and Management of Model Code
24. After approving a Model or Model change and releasing it for live trading, Two
Sigma stored the Model’s core code in a secure file known as the “Jar.”
25. To maintain control over its Models, Two Sigma limited Jar access to a select
group of engineering employees that were authorized to modify Model code.

7

26. Consequently, modelers, including Wu, could not access the Jar and, thus, could
modify a Model’s code only through the PAM process.  This restriction was intended to ensure
that all changes to Model code were subject to review and approval before the changes were
implemented into Models that had already been released for live trading.
27. By no later than 2019, Two Sigma’s Models grew in complexity to the point that
they exceeded the Jar’s storage capacity, leading the firm to store certain Model “parameters” in
a separate database called “celFS.”   Parameters are variable inputs that impact a Model’s
forecasts of a security’s performance.
28. Some Two Sigma modelers, including Wu, stored parameters for their Models in
celFS and then linked those parameters to the Model code stored in the Jar, enabling the Model
to operate using parameters stored outside the Jar.
29. One critical parameter that some modelers, including Wu, stored in celFS was the
“Decorrelation Parameter,” which sought to adjust a Model’s correlation to other Models and
Consolidated Forecasts.  As a general matter, the Decorrelation Parameter was inversely related
to the expected correlation:  as the parameter’s value decreased, a Model’s expected correlation
to other Models and Consolidated Forecasts increased, and vice versa.
30. Two Sigma used Decorrelation Parameters to ensure that, consistent with the
PAM process, each Model remained sufficiently uncorrelated from other Models and was likely
to generate alpha not already captured by those other Models.
31. For Two Sigma, the Decorrelation Parameter was important to preventing
redundancy among its Models, which could lead to unintended trading activity, such as the firm
buying or selling securities on behalf of its clients in different amounts, concentrations, and
frequencies than it intended.

8

32. During the Relevant Period, celFS had vulnerabilities that enabled modelers to
make unauthorized changes to Model parameters stored in celFS which, in turn, altered the
securities forecasts those Models generated.
33. Before June 2022, although prohibited by Two Sigma’s policies and procedures,
Two Sigma modelers could directly access and modify parameters in celFS, thus bypassing the
required PAM process.
34. In June 2022, Two Sigma introduced a ticket system for changing celFS
parameters, requiring modelers to submit written requests that were then implemented by Two
Sigma engineers.  However, because engineers implemented ticket system changes without
substantively reviewing the requests or ensuring management approval, modelers could also use
the ticket system to circumvent the required PAM process.
III. Wu Created Multiple Models for Two Sigma.
35. As a modeler at Two Sigma, Wu received training on the firm’s policies and
procedures for developing Models, created and submitted multiple Models for approval, and
understood the importance of decorrelation in the Model approval process.
A. Two Sigma Trained Wu on Model Development.
36. In April 2018, TSI hired Wu as a quantitative researcher to develop Models that
generated forecasts of securities’ future performance.
37. As part of his onboarding, Wu received training on Two Sigma’s Model
development policies and procedures, including the PAM process, which emphasized the
importance of decorrelation between Models.
38. Wu’s direct supervisor (“Supervisor A”) also provided Wu with guidance on
Model development, the PAM process, and Two Sigma’s policies and procedures.

9

39. Additionally, Two Sigma provided Wu with an employee handbook outlining the
firm’s policies and procedures, including t h e   PA M   process.
40. From 2019 to 2023, Wu annually certified that he had received, reviewed, and
understood specific “key documents,” including the employee handbook, and acknowledged his
responsibility to comply with all of Two Sigma’s policies and procedures.
41. Throughout the Relevant Period, Wu understood that decorrelation was a critical
aspect of Two Sigma’s model approval process, and that the firm would only approve a proposed
Model or Model change that did not exceed specific correlation thresholds relative to other
Models and the Consolidated Forecasts.
B. Wu Knew That His Compensation Was Tied to Model Performance.
42. Wu was aware that Two Sigma used his Models to buy and sell publicly-traded
equity securities, among other securities, on behalf of its advisory clients.
43. As Wu also knew, his annual incentive compensation, including cash bonuses and
performance grants, was tied to his Models’ performance, and his compensation was based in
part on his Models’ ability to generate unique investment returns that were uncorrelated to the
firm’s other Models and Consolidated Forecasts.
44. During the Relevant Period, Two Sigma provided Wu with access to periodic
reports concerning how each of his Models that were being used to trade securities for Two
Sigma’s clients were performing.  These reports included the profits and losses attributable to
each of his Models, as well as other metrics, which enabled Wu to monitor his Models’
performance on a near-daily basis.
45. Wu used these periodic reports to assess his Models’ performance and to justify
requests he made to his supervisors for increased   bonus compensation.

10

C. Wu Created At Least Fourteen Models for Two Sigma.
46. Between 2019 and 2023, Wu created or helped to create at least fourteen new
Models ( including Models A through N), which were submitted for approval through the PA M
process.
47. Wu’s Models generated predictions and trading signals for securities, including
primarily for buying and selling U.S. equity securities listed on exchanges such as the New York
Stock Exchange.
48. In designing these Models, Wu linked a Decorrelation Parameter stored in celFS
to each Model’s code and set the Decorrelation Parameter at a value necessary to satisfy Two
Sigma’s correlation tests required by the PAM process.
49. When Wu submitted his Models to Two Sigma for approval, he certified, among
other things, that he was providing the complete, accurate, and final version of each Model.
50. From August 2021 to December 2022, Wu submitted written attestations for at
least nine proposed new Models, affirming that each was “accurate and complete, including all
the information in the Model Inputs section” and that he had “followed research and simulation
best practices.”
51. For example, Wu submitted such attestations when seeking approval for Models
A, B, and C on or about April 20, 2021, October 25, 2021, and March 4, 2022, respectively.
52. At Two Sigma, “  simulation best practices” meant that a modeler was submitting
their best and final proposed Model for review, without intending to make further changes.
53. Starting in March 2022, Wu also submitted written attestations for at least two
proposed Models affirming that each was the “finalized version of the Model, [and is] exactly

11

how the [M]odel will behave [if it is approved and released for trading]. . . .  This includes (but is
not limited to): . . . Model decorrelation.”
54. Wu submitted these attestations when seeking approval for Models C and E on or
about March 4, 2022 and May 9, 2022, respectively.
55. However, Wu’s attestations were false when made, as he intended to, and did,
alter the Models from the versions submitted for approval by changing their Decorrelation
Parameters.
IV. Wu Engaged in a Scheme to Manipulate His Models.
56. Wu manipulated at least fourteen of his Models by secretly reducing the
Decorrelation Parameters stored in celFS that were submitted to and approved by Two Sigma
management.  This manipulation caused his Models to essentially replicate the outputs of Two
Sigma’s other Models and Consolidated Forecasts.  As a result, Wu deceived Two Sigma about
his Models’ true characteristics, caused Two Sigma to engage in securities trades that harmed
certain advisory clients, and enriched himself by millions of dollars.
A. Wu Secretly Changed the Decorrelation Parameters for His Models.
57. Beginning in November 2021, Wu secretly altered the Decorrelation Parameters
for at least fourteen of his Models, reducing their decorrelation with (and thus increasing their
correlation to) Two Sigma’s other Models and Consolidated Forecasts.
58. As Two Sigma modelers, including Wu, understood from information they
received from the firm, any single Model was unlikely to outperform the firm’s Consolidated
Forecasts.  By reducing the Decorrelation Parameters, Wu more closely aligned his Models’
performance with the Consolidated Forecasts, creating the false impression of his Models’ true
performance and contributions to the profits and losses in Two Sigma’s clients’ accounts.

12

59. Wu made these unauthorized changes to his Models without Two Sigma’s
knowledge or approval, circumventing the PAM process.  He often did so by submitting a Model
version with a specific Decorrelation Parameter in the PAM process for Two Sigma’s review,
only to secretly modify the Decorrelation Parameter in celFS to a significantly lower value than
he had represented, around the same time.
60. As a result of Wu’s unauthorized changes, his Models more closely mirrored
forecasts generated by other Models and Consolidated Forecasts, rather than generating unique
forecasts.  This increased the risk that Two Sigma would make unintended investment decisions,
such as over-concentration in, or over-exposure to, particular securities.
61. Before June 2022, Wu altered Decorrelation Parameters directly in celFS for at
least ten Models, without disclosure to or approval from Two Sigma.  In fact, Wu reduced these
Decorrelation Parameters to zero or near zero, effectively disabling Two Sigma’s safeguard for
preventing his Models from mirroring other Models and Consolidated Forecasts.
62. For example, Wu secretly changed the Decorrelation Parameter for Model A on
two occasions: on November 15, 2021, he reduced the parameter to less than one percent of its
approved value; on December 25, 2021, he reduced the parameter to zero.    On both occasions,
Wu made these changes directly in celFS without Two Sigma’s knowledge or approval.
63. After June 2022, Wu used the ticket system to make unauthorized changes to the
Decorrelation Parameter for four other Models, secretly reducing them to nearly zero by
submitting ticket requests to replace the celFS files that were part of his requests for Model
approval with new celFS files containing much smaller Decorrelation Parameter values.
64. For example, the below chart illustrates Wu’s manipulation of the Decorrelation
Parameters for Models B, F, and H.  In each case, Wu represented one Decorrelation Parameter

13

to Two Sigma for approval, but then secretly added two zeros to the parameter value in a new
celFS file, effectively reducing those parameters by a factor of 100:

Model
Decorrelation Parameter
Submitted to PAM Process
Wu’s Unauthorized
Modified Parameter
B 0.00661221 0.0000661221
F 0.00473471 0.0000473471
H 0.00661221 0.0000661221
65. By making these kinds of visually subtle (but quantitatively significant)
adjustments to the parameters, Wu attempted to conceal his actions, underscoring the deceptive
nature of his scheme.
66. Wu’s manipulation of the Decorrelation Parameters rendered his statements to
Two Sigma about his Models false and misleading; he represented that the versions of his
Models that were submitted for approval were complete and accurate when, in fact, he intended
to, and did, change the Decorrelation Parameters without disclosing those changes or obtaining
approval for them.
B. Wu Benefitted From—and Then Attempted to Hide—His Scheme.
67. Wu’s secret and deceptive modifications to the Decorrelation Parameters for his
Models were motivated by a desire to increase his incentive compensation, which was directly
tied to the performance of his Models in generating profits for Two Sigma’s clients.
68. By manipulating the Decorrelation Parameters, Wu artificially inflated the
apparent contribution of his Models to Two Sigma’s clients’ profits.  Two Sigma then paid Wu
millions of dollars of compensation based on these inflated contributions.
69. Following Wu’s manipulation of Decorrelation Parameters beginning in
November 2021, his incentive compensation for 2022 increased substantially.  TSI awarded Wu

14

more than $23 million for 2022, which consisted of a $16 million cash bonus and more than
$7 million in performance grants.
70. Indeed, Wu’s 2022 compensation exceeded three times the total compensation he
had received since joining TSI in April 2018.
71. The below chart summarizes Wu’s compensation by year:
Year Salary Cash Bonus Performance Grants Total
2018 $128,077 $286,678 $0 $414,755
2019 $193,000 $307,000 $0 $500,000
2020 $200,000 $2,600,000 $1,400,000 $4,200,000
2021 $225,000 $1,775,000 $750,000 $2,750,000
2022 $259,000 $16,000,000 $7,250,000 $23,509,000

72. TSI was unaware of Wu’s unauthorized changes to Decorrelation Parameters
when it issued his incentive compensation for the years 2021 and 2022.  Although TSI ultimately
cancelled Wu’s performance grants for these years after his termination, it has not recovered any
portion of the nearly $18 million in cash bonuses it paid Wu for those years.
73. Wu’s scheme began to unravel in January 2023, when TSI discovered an
anonymous social media post revealing that a TSI employee had received over $23 million in
compensation for 2022, prompting an investigation into Wu’s Models.
74. As TSI investigated, it found that Wu’s Models exhibited unusually high
correlations to other Models and Consolidated Forecasts.  When questioned about his Models’
performance, Wu failed to disclose the unauthorized changes he had made to the Decorrelation
Parameters, instead blaming the high correlations on other factors, such as market conditions or
the performance of other Models.

15

75. Upon learning of TSI’s investigation, Wu attempted to conceal his wrongdoing by
making additional unauthorized changes to the Decorrelation Parameters for his Models in order
to reverse his earlier changes.
76. For example, on July 29, 2023, Wu reverted the Decorrelation Parameters for
Models A, B, D, and E to the values that Two Sigma had originally approved in 2021 or 2022.
77. In August 2023, Wu admitted to Supervisor A and other TSI employees that he
had reduced the Decorrelation Parameters without approval, leading to his placement on
administrative leave and subsequent termination in January 2024 (which was effective as of
August 2023).
78. TSI’s investigation concluded that Wu’s manipulation of the Decorrelation
Parameters created a false impression that his Models generated more unique and profitable
investment returns than they actually did, resulting in Two Sigma making unintended purchases
and sales of securities on behalf of its clients and causing harm to many clients.
79. In December 2023 and January 2024, Two Sigma voluntarily repaid its negatively
impacted client funds and SMAs, which primarily included outside investors, approximately
$165 million.
FIRST CLAIM FOR RELIEF
Violations of Securities Act Section 17(a)
80. The SEC re-alleges and incorporates by reference here the allegations in
paragraphs 1 through 79.
81. Wu, directly or indirectly, singly or in concert, in the offer or sale of securities and
by the use of the means or instruments of transportation or communication in interstate
commerce or the mails, (i) knowingly or recklessly has employed one or more devices, schemes

16

or artifices to defraud, (ii) knowingly, recklessly, or negligently has obtained money or property
by means of one or more untrue statements of a material fact or omissions of a material fact
necessary in order to make the statements made, in light of the circumstances under which they
were made, not misleading, and/or (iii) knowingly, recklessly, or negligently has engaged in one
or more transactions, practices, or courses of business which operated or would operate as a
fraud or deceit upon the purchaser.
82. By reason of the foregoing, Wu, directly or indirectly, singly or in concert, has
violated and, unless enjoined, will again violate Securities Act Section 17(a)(1)-(3) [15 U.S.C.
§§ 77q(a)(1)-(3)].
SECOND CLAIM FOR RELIEF
Violations of Exchange Act Section 10(b) and Rule 10b-5   Thereunder
83. The SEC re-alleges and incorporates by reference here the allegations in
paragraphs 1 through 79.
84. Wu, directly or indirectly, singly or in concert, in connection with the purchase or
sale of securities and by the use of means or instrumentalities of interstate commerce, or the
mails, or the facilities of a national securities exchange, knowingly or recklessly has
(i) employed one or more devices, schemes, or artifices to defraud, (ii) made one or more untrue
statements of a material fact or omitted to state one or more material facts necessary in order to
make the statements made, in light of the circumstances under which they were made, not
misleading, and/or (iii) engaged in one or more acts, practices, or courses of business which
operated or would operate as a fraud or deceit upon other persons.

17

85. By reason of the foregoing, Wu, directly or indirectly, singly or in concert, has
violated and, unless enjoined, will again violate Exchange Act Section 10(b) [15 U.S.C. § 78j(b)]
and Rule 10b-5(a)-(c) thereunder [17 C.F.R. § 240.10b-5(a)-(c)].
PRAYER FOR RELIEF
WHEREFORE, the Commission respectfully requests that the Court enter a Final
Judgment:
I.
Permanently enjoining Wu and his agents, servants, employees and attorneys and all
persons in active concert or participation with him, from violating, directly or indirectly,
Securities Act Section 17(a) [15 U.S.C. § 77q(a)] and Exchange Act Section 10(b) [15 U.S.C. §
78j(b)] and Rule 10b-5 thereunder [17 C.F.R. § 240.10b-5] by committing or engaging in
specified actions or activities relevant to such violations;
II.
Ordering Wu to disgorge all ill-gotten gains he received directly or indirectly, with pre-
judgment interest thereon, as a result of the alleged violations, under Exchange Act Sections
21(d)(3), 21(d)(5), and 21(d)(7) [15 U.S.C. §§ 78u(d)(3), 78u(d)(5), and 78u(d)(7)];
III.
Ordering Wu to pay civil monetary penalties under Securities Act Section 20(d) [15
U.S.C. § 77t(d)] and Exchange Act Section 21(d)(3) [15 U.S.C. § 78u(d)(3)];
I V.
Permanently prohibiting Wu from, directly or indirectly, acting as or being associated
with any investment adviser, under Securities Act Section 20(e) [15 U.S.C. § 77t(e)] and
Exchange Act Section 21(d)(2) [15 U.S.C. § 78u(d)(2)].  For purposes of this paragraph, a person

18

is associated with an investment adviser if such person is a partner, officer, or director of such
investment adviser (or performs similar functions), or directly or indirectly controls or is
controlled by such investment adviser, including any employee of such investment adviser; and
V.
Granting any other and further relief this Court may deem just and proper.
JURY DEMAND
 The Commission demands a trial by jury.
Dated:  New York, New York
September 11, 2025
/s/ Christopher M. Colorado
Lee A. Greenwood
Christopher M. Colorado
David F. Benson*
Brian A. Kudon
Attorneys for Plaintiff
SECURITIES AND EXCHANGE
COMMISSION
New York Regional Office
100 Pearl Street
Suite 20-100
New York, NY 10004-2616
212-336-9143 (Colorado)
[email protected]

 *Pro Hac Vice Application Forthcoming

1 

Lee A. Greenwood 
Christopher M. Colorado 
Brian A. Kudon 
David F. Benson* 
Attorneys for Plaintiff 
SECURITIES AND EXCHANGE COMMISSION 
New York Regional Office 
100 Pearl Street, Suite 20-100 
New York, NY  10004-2616 
212-336-9143 (Colorado)
[email protected]

*Pro Hac Vice Application Forthcoming

UNITED STATES DISTRICT COURT 
SOUTHERN DISTRICT OF NEW YORK 

SECURITIES AND EXCHANGE COMMISSION, 

Plaintiff, 

-against-

JIAN WU, 

Defendant. 

COMPLAINT 

1:25 Civ. 07573

JURY TRIAL DEMANDED 

Plaintiff Securities and Exchange Commission (“Commission”), for its Complaint against 

Defendant Jian Wu (“Wu”), alleges: 

INTRODUCTION 

1. From 2021 to 2023, Jian Wu, a former employee of Two Sigma Investments, LP

(“TSI”), orchestrated a scheme to deceive his employer and reap millions of dollars in ill-gotten 

gains by manipulating computer-based algorithmic investment models (“Models”) that TSI and 

its affiliate, Two Sigma Advisers, LP (“TSA,” and, together, “Two Sigma”), used to make 

investment decisions for their clients, including decisions to buy and sell securities. 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 1 of 18



2 
 
 

2. Two Sigma developed and relied on Models to generate predictions and trading 

signals for securities prices, which it aggregated into “Consolidated Forecasts” that sought to 

predict the future performance of specific securities.  Two Sigma then relied on Models and 

Consolidated Forecasts to decide whether to buy and sell securities for its advisory clients. 

3. Two Sigma required that each new Model it developed and approved be 

sufficiently uncorrelated to, or “decorrelated” from, existing Models and Consolidated Forecasts 

to ensure that new Models generated unique forecasts that complimented, rather than duplicated, 

the firm’s existing predictions.  Two Sigma thus sought to increase the likelihood that new 

Models generated “alpha”—investment returns relative to the return of a relevant market 

benchmark—that was not already captured by its existing Models and Consolidated Forecasts. 

4. In violation of Two Sigma’s policies, between November 2021 and August 2023 

(the “Relevant Period”), Wu secretly manipulated at least fourteen Models he created or helped 

create, making unauthorized and undisclosed changes to those Models to increase their 

correlation to the firm’s other Models and Consolidated Forecasts.  Thus, instead of generating 

the unique trading signals and forecasts that Two Sigma expected, Wu’s manipulated Models 

effectively replicated the output of other Models and Consolidated Forecasts. 

5. Wu misrepresented to Two Sigma that his Models fell below the firm’s applicable 

correlation thresholds and were generating unique forecasts, deceiving Two Sigma into relying 

more heavily on them.  Consequently, when it deployed Wu’s Models, Two Sigma bought and 

sold securities for its clients in amounts, concentrations, and frequencies that differed from its 

intended strategies.  Wu’s secret Model changes caused certain Two Sigma clients to trade 

securities in a manner that caused harm to them of at least $165 million during the Relevant 

Period, which Two Sigma subsequently repaid. 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 2 of 18



3 
 
 

6. Wu’s scheme also caused Two Sigma to falsely believe that his Models were 

generating substantially more alpha not captured by the firm’s other Models and Consolidated 

Forecasts than his Models actually generated, which caused Two Sigma to pay Wu millions of 

dollars in cash bonuses and performance grants during the Relevant Period that it would not 

otherwise have paid him.  

7. Wu’s unauthorized and undisclosed changes to his Models and misrepresentations 

to Two Sigma about his Models’ correlations violated the antifraud provisions of the Securities 

Act of 1933 (“Securities Act”) and the Securities Exchange Act of 1934 (“Exchange Act”). 

NATURE OF THE PROCEEDINGS AND RELIEF SOUGHT 

8. The Commission brings this action under the authority conferred upon it by 

Securities Act Sections 20(b) and 20(d) [15 U.S.C. §§ 77t(b) and (d)] and Exchange Act Section 

21(d) [15 U.S.C. §§ 78u(d)]. 

9. The Commission seeks a final judgment:  (a) permanently enjoining Wu from 

violating the federal securities laws and rules this Complaint alleges he has violated; (b) ordering 

Wu to disgorge all ill-gotten gains he received as a result of the violations alleged here and to pay 

prejudgment interest thereon, under Exchange Act Sections 21(d)(3), 21(d)(5), and 21(d)(7) 

[15 U.S.C. §§ 78u(d)(3), 78u(d)(5), and 78u(d)(7)]; (c) ordering Wu to pay civil money penalties 

under Securities Act Section 20(d) [15 U.S.C. § 77t(d)] and Exchange Act Section 21(d)(3) [15 

U.S.C. § 78u(d)(3)]; (d) permanently prohibiting Wu from, directly or indirectly, acting as, or 

being associated, with any investment adviser, under Securities Act Section 20(e) [15 U.S.C. 

§ 77t(e)] and Exchange Act Section 21(d)(2) [15 U.S.C. § 78u(d)(2)]; and (e) ordering any other 

and further relief the Court may deem just and proper. 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 3 of 18



4 
 
 

JURISDICTION AND VENUE 

10. This Court has jurisdiction over this action under Securities Act Section 22(a) 

[15 U.S.C. § 77v(a)] and Exchange Act Section 27 [15 U.S.C. § 78aa]. 

11. Wu, directly and indirectly, has made use of the means or instrumentalities of 

interstate commerce or of the mails in connection with the transactions, acts, practices, and 

courses of business alleged herein. 

12. Venue lies in this District under Securities Act Section 22(a) [15 U.S.C. § 77v(a)] 

and Exchange Act Section 27 [15 U.S.C. § 78aa].  Wu transacted business in this District during 

the Relevant Period.  In addition, certain of the transactions, acts, practices, and courses of 

business alleged in this Complaint occurred within this District, including because Two Sigma, 

where Wu was employed and to which Wu made materially false and misleading statements, had 

offices located in this District to which Wu was assigned. 

DEFENDANT 

13. Jian Wu, age 34, is a permanent legal resident of New York, NY, and a citizen of 

China.  He holds Ph.D.’s in Industrial Engineering and Operations Research from the University 

of Southern California and Cornell University, respectively.  Wu was employed by TSI from 

April 2018 to August 2023, holding various positions, including quantitative researcher, vice 

president, and senior vice president. 

RELEVANT NON-PARTY ENTITIES 

14. TSI is a Delaware limited partnership headquartered in New York, NY.  TSI was 

founded in July 2001 and has been registered with the Commission as an investment adviser since 

August 2009.  TSI provides advisory services on a discretionary basis to various clients, including 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 4 of 18



5 
 
 

private investment funds.  According to its Form ADV dated July 25, 2025, TSI had regulatory 

assets under management of more than $110 billion. 

15. TSA is a Delaware limited partnership headquartered in New York, NY.  TSA was 

founded in December 2001 and has been registered with the Commission as an investment adviser 

since February 2010.  TSA provides advisory services on a discretionary basis to various clients, 

including private investment funds, a registered investment company, foreign funds, and separately 

managed accounts (“SMAs”).  According to its Form ADV dated July 25, 2025, TSA had regulatory 

assets under management of more than $74 billion. 

STATEMENT OF FACTS 

I. Two Sigma’s Model Development and Approval Process 

16. During the Relevant Period, TSI developed Models and then licensed those 

Models to TSA.  Both TSI and TSA used Models and Consolidated Forecasts to make investment 

decisions for their clients, including buying and selling equity securities listed on exchanges such 

as the New York Stock Exchange. 

17. Two Sigma required new Models and changes to existing Models to be 

sufficiently distinct from its other Models and Consolidated Forecasts, such that new or modified 

Models were likely to generate alpha not already captured by the firm’s other Models or 

Consolidated Forecasts. 

18. During the Relevant Period, Two Sigma’s process for developing and approving 

new Models or Model changes was described in its “Productionalize a Model” (“PAM”) manual, 

which set forth the steps Two Sigma required before approving a new Model or Model change 

(referred to collectively as the “PAM process”). 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 5 of 18



6 
 
 

19. As part of the PAM process, Two Sigma required modelers to submit Model 

proposals or proposed Model changes through an internal system called “TOM.”  These required 

submissions included detailed information about the Model proposal, such as its expected 

correlation to existing Models and Consolidated Forecasts based on simulations run on historical 

securities trading data. 

20. Two Sigma also required the lead modeler for each Model to certify that the 

information submitted with the TOM proposal, including all Model inputs, was accurate and 

complete. 

21. Two Sigma management then reviewed the TOM submissions and generally 

approved only new Models or Model changes that met specific criteria, including a sufficient 

lack of correlation to the firm’s existing Models and Consolidated Forecasts. 

22. Once a new Model or Model change was approved, Two Sigma’s portfolio 

management teams determined how to incorporate the Model into the firm’s investment 

strategies for their advisory clients. 

23. By following the PAM process, Two Sigma aimed to maintain an uncorrelated set 

of Models that would generate unique investment returns, thereby enhancing the overall 

performance of its advisory clients’ investments. 

II. Two Sigma’s Storage and Management of Model Code 

24. After approving a Model or Model change and releasing it for live trading, Two 

Sigma stored the Model’s core code in a secure file known as the “Jar.”  

25. To maintain control over its Models, Two Sigma limited Jar access to a select 

group of engineering employees that were authorized to modify Model code. 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 6 of 18



7 
 
 

26. Consequently, modelers, including Wu, could not access the Jar and, thus, could 

modify a Model’s code only through the PAM process.  This restriction was intended to ensure 

that all changes to Model code were subject to review and approval before the changes were 

implemented into Models that had already been released for live trading. 

27. By no later than 2019, Two Sigma’s Models grew in complexity to the point that 

they exceeded the Jar’s storage capacity, leading the firm to store certain Model “parameters” in 

a separate database called “celFS.”  Parameters are variable inputs that impact a Model’s 

forecasts of a security’s performance. 

28. Some Two Sigma modelers, including Wu, stored parameters for their Models in 

celFS and then linked those parameters to the Model code stored in the Jar, enabling the Model 

to operate using parameters stored outside the Jar.   

29. One critical parameter that some modelers, including Wu, stored in celFS was the 

“Decorrelation Parameter,” which sought to adjust a Model’s correlation to other Models and 

Consolidated Forecasts.  As a general matter, the Decorrelation Parameter was inversely related 

to the expected correlation:  as the parameter’s value decreased, a Model’s expected correlation 

to other Models and Consolidated Forecasts increased, and vice versa. 

30. Two Sigma used Decorrelation Parameters to ensure that, consistent with the 

PAM process, each Model remained sufficiently uncorrelated from other Models and was likely 

to generate alpha not already captured by those other Models. 

31. For Two Sigma, the Decorrelation Parameter was important to preventing 

redundancy among its Models, which could lead to unintended trading activity, such as the firm 

buying or selling securities on behalf of its clients in different amounts, concentrations, and 

frequencies than it intended. 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 7 of 18



8 
 
 

32. During the Relevant Period, celFS had vulnerabilities that enabled modelers to 

make unauthorized changes to Model parameters stored in celFS which, in turn, altered the 

securities forecasts those Models generated. 

33. Before June 2022, although prohibited by Two Sigma’s policies and procedures, 

Two Sigma modelers could directly access and modify parameters in celFS, thus bypassing the 

required PAM process.   

34. In June 2022, Two Sigma introduced a ticket system for changing celFS 

parameters, requiring modelers to submit written requests that were then implemented by Two 

Sigma engineers.  However, because engineers implemented ticket system changes without 

substantively reviewing the requests or ensuring management approval, modelers could also use 

the ticket system to circumvent the required PAM process. 

III. Wu Created Multiple Models for Two Sigma. 

35. As a modeler at Two Sigma, Wu received training on the firm’s policies and 

procedures for developing Models, created and submitted multiple Models for approval, and 

understood the importance of decorrelation in the Model approval process. 

A. Two Sigma Trained Wu on Model Development. 

36. In April 2018, TSI hired Wu as a quantitative researcher to develop Models that 

generated forecasts of securities’ future performance. 

37. As part of his onboarding, Wu received training on Two Sigma’s Model 

development policies and procedures, including the PAM process, which emphasized the 

importance of decorrelation between Models. 

38. Wu’s direct supervisor (“Supervisor A”) also provided Wu with guidance on 

Model development, the PAM process, and Two Sigma’s policies and procedures. 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 8 of 18



9 
 
 

39. Additionally, Two Sigma provided Wu with an employee handbook outlining the 

firm’s policies and procedures, including the PAM process.   

40. From 2019 to 2023, Wu annually certified that he had received, reviewed, and 

understood specific “key documents,” including the employee handbook, and acknowledged his 

responsibility to comply with all of Two Sigma’s policies and procedures. 

41. Throughout the Relevant Period, Wu understood that decorrelation was a critical 

aspect of Two Sigma’s model approval process, and that the firm would only approve a proposed 

Model or Model change that did not exceed specific correlation thresholds relative to other 

Models and the Consolidated Forecasts. 

B. Wu Knew That His Compensation Was Tied to Model Performance. 

42. Wu was aware that Two Sigma used his Models to buy and sell publicly-traded 

equity securities, among other securities, on behalf of its advisory clients. 

43. As Wu also knew, his annual incentive compensation, including cash bonuses and 

performance grants, was tied to his Models’ performance, and his compensation was based in 

part on his Models’ ability to generate unique investment returns that were uncorrelated to the 

firm’s other Models and Consolidated Forecasts. 

44. During the Relevant Period, Two Sigma provided Wu with access to periodic 

reports concerning how each of his Models that were being used to trade securities for Two 

Sigma’s clients were performing.  These reports included the profits and losses attributable to 

each of his Models, as well as other metrics, which enabled Wu to monitor his Models’ 

performance on a near-daily basis. 

45. Wu used these periodic reports to assess his Models’ performance and to justify 

requests he made to his supervisors for increased bonus compensation. 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 9 of 18



10 
 
 

C. Wu Created At Least Fourteen Models for Two Sigma. 

46. Between 2019 and 2023, Wu created or helped to create at least fourteen new 

Models (including Models A through N), which were submitted for approval through the PAM 

process. 

47. Wu’s Models generated predictions and trading signals for securities, including 

primarily for buying and selling U.S. equity securities listed on exchanges such as the New York 

Stock Exchange. 

48. In designing these Models, Wu linked a Decorrelation Parameter stored in celFS 

to each Model’s code and set the Decorrelation Parameter at a value necessary to satisfy Two 

Sigma’s correlation tests required by the PAM process. 

49. When Wu submitted his Models to Two Sigma for approval, he certified, among 

other things, that he was providing the complete, accurate, and final version of each Model. 

50. From August 2021 to December 2022, Wu submitted written attestations for at 

least nine proposed new Models, affirming that each was “accurate and complete, including all 

the information in the Model Inputs section” and that he had “followed research and simulation 

best practices.” 

51. For example, Wu submitted such attestations when seeking approval for Models 

A, B, and C on or about April 20, 2021, October 25, 2021, and March 4, 2022, respectively. 

52. At Two Sigma, “simulation best practices” meant that a modeler was submitting 

their best and final proposed Model for review, without intending to make further changes. 

53. Starting in March 2022, Wu also submitted written attestations for at least two 

proposed Models affirming that each was the “finalized version of the Model, [and is] exactly 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 10 of 18



11 
 
 

how the [M]odel will behave [if it is approved and released for trading]. . . .  This includes (but is 

not limited to): . . . Model decorrelation.” 

54. Wu submitted these attestations when seeking approval for Models C and E on or 

about March 4, 2022 and May 9, 2022, respectively. 

55. However, Wu’s attestations were false when made, as he intended to, and did, 

alter the Models from the versions submitted for approval by changing their Decorrelation 

Parameters. 

IV. Wu Engaged in a Scheme to Manipulate His Models. 

56. Wu manipulated at least fourteen of his Models by secretly reducing the 

Decorrelation Parameters stored in celFS that were submitted to and approved by Two Sigma 

management.  This manipulation caused his Models to essentially replicate the outputs of Two 

Sigma’s other Models and Consolidated Forecasts.  As a result, Wu deceived Two Sigma about 

his Models’ true characteristics, caused Two Sigma to engage in securities trades that harmed 

certain advisory clients, and enriched himself by millions of dollars. 

A. Wu Secretly Changed the Decorrelation Parameters for His Models. 

57. Beginning in November 2021, Wu secretly altered the Decorrelation Parameters 

for at least fourteen of his Models, reducing their decorrelation with (and thus increasing their 

correlation to) Two Sigma’s other Models and Consolidated Forecasts. 

58. As Two Sigma modelers, including Wu, understood from information they 

received from the firm, any single Model was unlikely to outperform the firm’s Consolidated 

Forecasts.  By reducing the Decorrelation Parameters, Wu more closely aligned his Models’ 

performance with the Consolidated Forecasts, creating the false impression of his Models’ true 

performance and contributions to the profits and losses in Two Sigma’s clients’ accounts. 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 11 of 18



12 
 
 

59. Wu made these unauthorized changes to his Models without Two Sigma’s 

knowledge or approval, circumventing the PAM process.  He often did so by submitting a Model 

version with a specific Decorrelation Parameter in the PAM process for Two Sigma’s review, 

only to secretly modify the Decorrelation Parameter in celFS to a significantly lower value than 

he had represented, around the same time. 

60. As a result of Wu’s unauthorized changes, his Models more closely mirrored 

forecasts generated by other Models and Consolidated Forecasts, rather than generating unique 

forecasts.  This increased the risk that Two Sigma would make unintended investment decisions, 

such as over-concentration in, or over-exposure to, particular securities. 

61. Before June 2022, Wu altered Decorrelation Parameters directly in celFS for at 

least ten Models, without disclosure to or approval from Two Sigma.  In fact, Wu reduced these 

Decorrelation Parameters to zero or near zero, effectively disabling Two Sigma’s safeguard for 

preventing his Models from mirroring other Models and Consolidated Forecasts. 

62. For example, Wu secretly changed the Decorrelation Parameter for Model A on 

two occasions: on November 15, 2021, he reduced the parameter to less than one percent of its 

approved value; on December 25, 2021, he reduced the parameter to zero.  On both occasions, 

Wu made these changes directly in celFS without Two Sigma’s knowledge or approval.   

63. After June 2022, Wu used the ticket system to make unauthorized changes to the 

Decorrelation Parameter for four other Models, secretly reducing them to nearly zero by 

submitting ticket requests to replace the celFS files that were part of his requests for Model 

approval with new celFS files containing much smaller Decorrelation Parameter values. 

64. For example, the below chart illustrates Wu’s manipulation of the Decorrelation 

Parameters for Models B, F, and H.  In each case, Wu represented one Decorrelation Parameter 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 12 of 18



13 
 
 

to Two Sigma for approval, but then secretly added two zeros to the parameter value in a new 

celFS file, effectively reducing those parameters by a factor of 100:  

 
Model 

Decorrelation Parameter  
Submitted to PAM Process 

Wu’s Unauthorized  
Modified Parameter 

B 0.00661221 0.0000661221 

F 0.00473471 0.0000473471 

H 0.00661221 0.0000661221 

65. By making these kinds of visually subtle (but quantitatively significant) 

adjustments to the parameters, Wu attempted to conceal his actions, underscoring the deceptive 

nature of his scheme. 

66. Wu’s manipulation of the Decorrelation Parameters rendered his statements to 

Two Sigma about his Models false and misleading; he represented that the versions of his 

Models that were submitted for approval were complete and accurate when, in fact, he intended 

to, and did, change the Decorrelation Parameters without disclosing those changes or obtaining 

approval for them. 

B. Wu Benefitted From—and Then Attempted to Hide—His Scheme. 

67. Wu’s secret and deceptive modifications to the Decorrelation Parameters for his 

Models were motivated by a desire to increase his incentive compensation, which was directly 

tied to the performance of his Models in generating profits for Two Sigma’s clients. 

68. By manipulating the Decorrelation Parameters, Wu artificially inflated the 

apparent contribution of his Models to Two Sigma’s clients’ profits.  Two Sigma then paid Wu 

millions of dollars of compensation based on these inflated contributions. 

69. Following Wu’s manipulation of Decorrelation Parameters beginning in 

November 2021, his incentive compensation for 2022 increased substantially.  TSI awarded Wu 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 13 of 18



14 
 
 

more than $23 million for 2022, which consisted of a $16 million cash bonus and more than 

$7 million in performance grants.   

70. Indeed, Wu’s 2022 compensation exceeded three times the total compensation he 

had received since joining TSI in April 2018. 

71. The below chart summarizes Wu’s compensation by year: 

Year Salary Cash Bonus Performance Grants Total 

2018 $128,077 $286,678 $0 $414,755 

2019 $193,000 $307,000 $0 $500,000 

2020 $200,000 $2,600,000 $1,400,000 $4,200,000 

2021 $225,000 $1,775,000 $750,000 $2,750,000 

2022 $259,000 $16,000,000 $7,250,000 $23,509,000 
 

72. TSI was unaware of Wu’s unauthorized changes to Decorrelation Parameters 

when it issued his incentive compensation for the years 2021 and 2022.  Although TSI ultimately 

cancelled Wu’s performance grants for these years after his termination, it has not recovered any 

portion of the nearly $18 million in cash bonuses it paid Wu for those years. 

73. Wu’s scheme began to unravel in January 2023, when TSI discovered an 

anonymous social media post revealing that a TSI employee had received over $23 million in 

compensation for 2022, prompting an investigation into Wu’s Models. 

74. As TSI investigated, it found that Wu’s Models exhibited unusually high 

correlations to other Models and Consolidated Forecasts.  When questioned about his Models’ 

performance, Wu failed to disclose the unauthorized changes he had made to the Decorrelation 

Parameters, instead blaming the high correlations on other factors, such as market conditions or 

the performance of other Models. 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 14 of 18



15 
 
 

75. Upon learning of TSI’s investigation, Wu attempted to conceal his wrongdoing by 

making additional unauthorized changes to the Decorrelation Parameters for his Models in order 

to reverse his earlier changes.   

76. For example, on July 29, 2023, Wu reverted the Decorrelation Parameters for 

Models A, B, D, and E to the values that Two Sigma had originally approved in 2021 or 2022. 

77. In August 2023, Wu admitted to Supervisor A and other TSI employees that he 

had reduced the Decorrelation Parameters without approval, leading to his placement on 

administrative leave and subsequent termination in January 2024 (which was effective as of 

August 2023). 

78. TSI’s investigation concluded that Wu’s manipulation of the Decorrelation 

Parameters created a false impression that his Models generated more unique and profitable 

investment returns than they actually did, resulting in Two Sigma making unintended purchases 

and sales of securities on behalf of its clients and causing harm to many clients. 

79. In December 2023 and January 2024, Two Sigma voluntarily repaid its negatively 

impacted client funds and SMAs, which primarily included outside investors, approximately 

$165 million. 

FIRST CLAIM FOR RELIEF 
Violations of Securities Act Section 17(a) 

80. The SEC re-alleges and incorporates by reference here the allegations in 

paragraphs 1 through 79. 

81. Wu, directly or indirectly, singly or in concert, in the offer or sale of securities and 

by the use of the means or instruments of transportation or communication in interstate 

commerce or the mails, (i) knowingly or recklessly has employed one or more devices, schemes 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 15 of 18



16 
 
 

or artifices to defraud, (ii) knowingly, recklessly, or negligently has obtained money or property 

by means of one or more untrue statements of a material fact or omissions of a material fact 

necessary in order to make the statements made, in light of the circumstances under which they 

were made, not misleading, and/or (iii) knowingly, recklessly, or negligently has engaged in one 

or more transactions, practices, or courses of business which operated or would operate as a 

fraud or deceit upon the purchaser. 

82. By reason of the foregoing, Wu, directly or indirectly, singly or in concert, has 

violated and, unless enjoined, will again violate Securities Act Section 17(a)(1)-(3) [15 U.S.C. 

§§ 77q(a)(1)-(3)]. 

SECOND CLAIM FOR RELIEF 
Violations of Exchange Act Section 10(b) and Rule 10b-5 Thereunder 

83. The SEC re-alleges and incorporates by reference here the allegations in 

paragraphs 1 through 79. 

84. Wu, directly or indirectly, singly or in concert, in connection with the purchase or 

sale of securities and by the use of means or instrumentalities of interstate commerce, or the 

mails, or the facilities of a national securities exchange, knowingly or recklessly has 

(i) employed one or more devices, schemes, or artifices to defraud, (ii) made one or more untrue 

statements of a material fact or omitted to state one or more material facts necessary in order to 

make the statements made, in light of the circumstances under which they were made, not 

misleading, and/or (iii) engaged in one or more acts, practices, or courses of business which 

operated or would operate as a fraud or deceit upon other persons. 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 16 of 18



17 
 
 

85. By reason of the foregoing, Wu, directly or indirectly, singly or in concert, has 

violated and, unless enjoined, will again violate Exchange Act Section 10(b) [15 U.S.C. § 78j(b)] 

and Rule 10b-5(a)-(c) thereunder [17 C.F.R. § 240.10b-5(a)-(c)]. 

PRAYER FOR RELIEF 

WHEREFORE, the Commission respectfully requests that the Court enter a Final 

Judgment: 

I. 

Permanently enjoining Wu and his agents, servants, employees and attorneys and all 

persons in active concert or participation with him, from violating, directly or indirectly, 

Securities Act Section 17(a) [15 U.S.C. § 77q(a)] and Exchange Act Section 10(b) [15 U.S.C. § 

78j(b)] and Rule 10b-5 thereunder [17 C.F.R. § 240.10b-5] by committing or engaging in 

specified actions or activities relevant to such violations; 

II. 

Ordering Wu to disgorge all ill-gotten gains he received directly or indirectly, with pre-

judgment interest thereon, as a result of the alleged violations, under Exchange Act Sections 

21(d)(3), 21(d)(5), and 21(d)(7) [15 U.S.C. §§ 78u(d)(3), 78u(d)(5), and 78u(d)(7)]; 

III. 

Ordering Wu to pay civil monetary penalties under Securities Act Section 20(d) [15 

U.S.C. § 77t(d)] and Exchange Act Section 21(d)(3) [15 U.S.C. § 78u(d)(3)];  

IV. 

Permanently prohibiting Wu from, directly or indirectly, acting as or being associated 

with any investment adviser, under Securities Act Section 20(e) [15 U.S.C. § 77t(e)] and 

Exchange Act Section 21(d)(2) [15 U.S.C. § 78u(d)(2)].  For purposes of this paragraph, a person 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 17 of 18



18 
 
 

is associated with an investment adviser if such person is a partner, officer, or director of such 

investment adviser (or performs similar functions), or directly or indirectly controls or is 

controlled by such investment adviser, including any employee of such investment adviser; and 

V. 

Granting any other and further relief this Court may deem just and proper.  

JURY DEMAND 

 The Commission demands a trial by jury.  

Dated: New York, New York 
September 11, 2025 

/s/ Christopher M. Colorado  
Lee A. Greenwood 
Christopher M. Colorado 
David F. Benson* 
Brian A. Kudon 
Attorneys for Plaintiff 
SECURITIES AND EXCHANGE 
COMMISSION 
New York Regional Office 
100 Pearl Street  
Suite 20-100 
New York, NY 10004-2616 
212-336-9143 (Colorado) 
[email protected]  
 
 *Pro Hac Vice Application Forthcoming 

Case 1:25-cv-07573     Document 1     Filed 09/11/25     Page 18 of 18


	Lee A. Greenwood Christopher M. Colorado Brian A. Kudon David F. Benson* Attorneys for Plaintiff SECURITIES AND EXCHANGE COMMISSION New York Regional Office 100 Pearl Street, Suite 20-100 New York, NY  10004-2616 212-336-9143 (Colorado) ColoradoCh@sec...
	*Pro Hac Vice Application Forthcoming
	Plaintiff Securities and Exchange Commission (“Commission”), for its Complaint against Defendant Jian Wu (“Wu”), alleges:
	INTRODUCTION
	NATURE OF THE PROCEEDINGS AND RELIEF SOUGHT
	JURISDICTION AND VENUE
	DEFENDANT
	RELEVANT NON-PARTY ENTITIES
	STATEMENT OF FACTS
	I. Two Sigma’s Model Development and Approval Process
	II. Two Sigma’s Storage and Management of Model Code
	III. Wu Created Multiple Models for Two Sigma.
	A. Two Sigma Trained Wu on Model Development.
	B. Wu Knew That His Compensation Was Tied to Model Performance.
	C. Wu Created At Least Fourteen Models for Two Sigma.

	IV. Wu Engaged in a Scheme to Manipulate His Models.
	B. Wu Benefitted From—and Then Attempted to Hide—His Scheme.

	FIRST CLAIM FOR RELIEF Violations of Securities Act Section 17(a)
	SECOND CLAIM FOR RELIEF Violations of Exchange Act Section 10(b) and Rule 10b-5 Thereunder
	PRAYER FOR RELIEF
	I.
	II.
	III.
	IV.
	V.
	JURY DEMAND