Opinion
CIVIL ACTION NO. 16-5173
2021-06-21
NUANCE COMMUNICATIONS, INC., Plaintiff, v. INTERNATIONAL BUSINESS MACHINES CORP., Defendant.
Daniela Kirsztajn, Robert William Taylor, Jessica Falk, Kevin F. Meade, David Lender, Weil Gotshal & Manges LLP, New York, NY, David Greenbaum, Nuance Communication, Inc., Mahwah, NJ, for Plaintiff. Elinor Catherine Sutton, Kevin Samuel Reed, Hope DeLaney Skibitsky, Quinn Emanuel Urquhart & Sullivan, New York, NY, Florentina Dragulescu Field, Pro Hac Vice, Quinn Emmanuel Urquhart & Sullivan, LLP, Washington, DC, for Defendant.
Daniela Kirsztajn, Robert William Taylor, Jessica Falk, Kevin F. Meade, David Lender, Weil Gotshal & Manges LLP, New York, NY, David Greenbaum, Nuance Communication, Inc., Mahwah, NJ, for Plaintiff.
Elinor Catherine Sutton, Kevin Samuel Reed, Hope DeLaney Skibitsky, Quinn Emanuel Urquhart & Sullivan, New York, NY, Florentina Dragulescu Field, Pro Hac Vice, Quinn Emmanuel Urquhart & Sullivan, LLP, Washington, DC, for Defendant.
MEMORANDUM
EDUARDO C. ROBRENO, District Judge TABLE OF CONTENTS
I. INTRODUCTION ...358
II. PROCEDURAL HISTORY ...359
III. FINDINGS OF FACT ...360
A. The Parties ...360
B. DeepQA ...360
C. The Software License Agreement ...360
1. The Negotiations...361
2. The Writing...362
D. Performance ...363
1. IBM Creates Updates...363
2. IBM Withholds Updates...364
3. The Parties’ Communications...366
E. Updates and Products ...366
IV. CONCLUSIONS OF LAW AND DISCUSSION ...368
A. Breach of Contract ...368
1. Contract Interpretation...368
2. Breach...371
B. Statute of Limitations ...373
1. Willful Blindness...375
2. Continuing Breach...382
3. Equitable Estoppel...383
V. SUMMARY ...384
I. INTRODUCTION
This is a breach of contract case arising under New York law. But more than that, this case is a contemporary window into the brave new world of artificial intelligence ("AI") commercial applications. In 2011, Watson, a computer system embedding IBM-designed software technology DeepQA, defeated two human former champions in a real-time competition on national television on the popular game show Jeopardy!. DeepQA accepts natural language questions and searches a body of information, generates a range of hypotheses, ranks these hypotheses, and returns the hypothesis it has the most confidence correctly answers the question, all in no more than seconds. The public debut of this modern deus ex machina received world-wide attention. But to fully interact with humans and be useful, Watson needed to understand and process questions beyond the Jeopardy! domain, asked through human voice.
"Yet the concept of intelligent computers, advanced by the British mathematician Alan Turing in 1950, isn't new; nor is the term ‘artificial intelligence,’ first used at a research conference in 1956." David A. Shaywitz, They Think They're So Smart, Wall St. J., May 22, 2021, at C7. "What has changed is AI's power and reach, especially with the arrival of what is called ‘deep learning’—the capacity for powerful pattern recognition, with seemingly little human instruction required." Id.
IBM's search for sophisticated partners with whom it could develop applications for DeepQA beyond winning Jeopardy! forms the backdrop of this case.
Beginning in 2010, even before DeepQA's triumphant appearance in Jeopardy!, IBM and Nuance were discussing possible partnership opportunities to build technological solutions to business problems. These discussions ripened in late 2010 into a Software License Agreement ("SLA"). The SLA entitled Nuance to one copy of IBM Research Group's "Automatic Open-Domain Question Answering" software system (which embedded IBM's DeepQA technology ), Tools to create commercial applications based on this technology and Nuance's technology, and ten years of software "updates." In return, Nuance paid IBM $25 million.
Although IBM and Nuance entered into collaboration agreements prior to 2010, see, e.g., Reuters Staff, Nuance to Buy IBM Speech Patents, to Collaborate, Reuters, https://www.reuters.com/article/ibm-nuance/nuance-to-buy-ibm-speech-patents-to-collaborate-idUSN1548243220090115 (Jan. 15, 2009, 9:06 AM), these agreements focused on other technology that is not at issue in this case.
For clarity's sake, and because the parties refer to it as such, the Court will refer to the software system as "DeepQA" for the remainder of the memorandum.
Nuance contends that under the SLA, IBM was to supply it with updates created by all IBM groups (i.e., not just IBM Research Group, which created the technology). IBM argues to the contrary, i.e., that the delivery of updates was limited to only those developed by the Research Group. Crucial to the distinction is that under IBM's interpretation of the SLA, it was not obligated to deliver to Nuance the "blue-washed" code (created by the Software Group), a set of updates which Nuance claims made DeepQA commercially viable.
For the reasons set forth below, the Court finds that Nuance was entitled to receive one set of updates that would allow DeepQA to be used in a commercial setting (i.e., the blue-washed code), and while IBM delivered some updates to Nuance, IBM breached the agreement by failing to deliver the blue-washed code. However, Nuance was sufficiently aware of the significant possibility that IBM was breaching the agreement prior to June 30, 2014 (i.e., more than two years before Nuance filed suit). Given what Nuance knew at the time, Nuance avoided making the appropriate inquiry to IBM and was willfully blind. Under these circumstances, Nuance's claims are barred by the contractual limitations period.
II. PROCEDURAL HISTORY
Nuance filed suit on June 30, 2016, in the United States District Court for the Southern District of New York. Nuance brought three counts: declaratory judgment, breach of contract, and breach of the implied covenant of good faith and fair dealing. After completing discovery, the parties filed cross-motions for summary judgment. Judge Karas, who was then presiding over the case, denied Nuance's motion but granted IBM's motion in part and denied it in part. Nuance's claim for breach of the implied covenant of good faith and fair dealing was dismissed with prejudice because it was duplicative of the breach of contract claim.
Due to temporary docket congestion in the Southern District of New York, and with Judge Karas's consent, the undersigned was designated by the Committee on Intercircuit Assignments to preside over all subsequent proceedings in the case. The matter proceeded to a bench trial in February 2020 for four days in New York. Due to the COVID-19 pandemic, the balance of the trial was then conducted at the end of July 2020 in Philadelphia, with some witnesses testifying via Zoom. After the bench trial concluded, the parties submitted proposed findings of fact and conclusions of law and the Court heard closing arguments. While the outcome of the case turns on the Court's determination that Nuance was willfully blind, for the sake of completeness and as a necessary part of the Court's decision on willful blindness, the Court has included in its decision factual and legal determinations on the issues relating to contract interpretation and the breach of contract claim.
The Honorable Eduardo C. Robreno serves as a United States District Judge in the Eastern District of Pennsylvania.
At trial, witnesses on direct examination testified by way of declarations. These witnesses were then subjected to cross-examination, re-direct, and re-cross. Also, the deposition testimonies of certain witnesses not available at trial were admitted. The declarations and the deposition testimonies are part of the trial record.
This memorandum constitutes the Court's findings of fact and conclusions of law. See Fed. R. Civ. P. 52(a)(1).
III. FINDINGS OF FACT
A. The Parties
1. Nuance Communications is a Delaware corporation with a principal place of business in Burlington, Massachusetts. Compl. 3, ECF No. 6. Nuance is a provider of voice recognition and natural language understanding technology, and it offers artificial intelligence solutions in industries such as mobile, automotive, and healthcare. Petro Decl. 2, ECF No. 211. Of these, the healthcare industry is Nuance's largest business segment. Petro Decl. 2.
Forbes recently reported that Microsoft has agreed to acquire Nuance. See Joe Cornell, Microsoft Announces Acquisition of Nuance; Targets Completion in 2021, Forbes (Apr. 20, 2021, 11:47 AM), https://www.forbes.com/sites/joecornell/2021/04/20/microsoft-announces-acquisition-of-nuance-targets-completion-in-2021/. That acquisition has no bearing on this case.
2. International Business Machines ("IBM") Corporation is a New York corporation with a principal place of business in Armonk, New York. Compl. 3; Answer 2, ECF No. 9. IBM Research Group ("IBM Research") is a subdivision of IBM that develops and licenses technologies but does not develop commercial software products. Brown Decl. 9, ECF No. 216-5; Kelly Decl. 4, ECF No. 216-7; Reardon Decl. 2–3, ECF No. 216-2; Trial Tr. 597; Summ. J. Op. 3, ECF No. 150. IBM Software Group ("IBM Software") is a subdivision of IBM that develops and markets IBM's commercial software products. McQueeney Decl. 3, ECF No. 216-1; Trial Tr. 598; Summ. J. Op. 4. These subdivisions are not separate legal entities—IBM Corporation is the relevant legal entity and owns the technology at issue. Trial Tr. 332–33, 341, 597–98; July 28 Trial Tr. 8; JX.001.1.
Citations to "JX" refer to the joint exhibits filed by the parties. Citations to "PX" refer to the plaintiff's exhibits filed by Nuance. Citations to "DX" refer to the defendant's exhibits filed by IBM.
B. DeepQA
3. DeepQA is a question-answering software technology, developed by IBM Research in the 2000s, which automatically accepts natural language questions and searches a body of information, generates a range of hypotheses, ranks these hypotheses, and returns the hypothesis it has the most confidence correctly answers the question. Brown Decl. 2–3; Schnell Decl. 9, ECF No. 34; PX001.045. This experimental technology, which uses natural language and mimics how humans think, was put to the test when embedded in Watson, a system capable of winning the television game show Jeopardy!, a goal achieved in 2011. Kelly Decl. 4. IBM ultimately intended not only to win Jeopardy! but also to create technology that had applications in business. Trial Tr. 257, 674–75.
C. The Software License Agreement
4. On June 3, 2010, IBM and Nuance had a daylong meeting, where IBM presented several IBM technologies and opportunities to Nuance for investment and partnership. McCann Dep. 42–43, ECF No. 131-10; PX002. IBM's purpose for the meeting was to persuade Nuance to partner with IBM to build technology solutions in the healthcare space. PX002. IBM showcased the DeepQA technology at this meeting, and Nuance later expressed interest in licensing DeepQA. Bloom Decl. 2, ECF No. 212; Trial Tr. 351; McCann Dep. 50–51; DX117.
1. The Negotiations
5. Nuance was interested in DeepQA for its potential application to business solutions in the healthcare space, not for its ability to compete on Jeopardy!. Ricci Decl. 3–4, ECF No. 214; Petro Decl. 2; Trial Tr. 105; PX005; PX007. Knowing of Nuance's business interest, IBM presented DeepQA to Nuance as having potential applications beyond Jeopardy!; and, while it had no concrete plans at the time, IBM told Nuance that IBM was committed to investing in developing those applications. Bloom Decl. 5; Ricci Decl. 5; Petro Decl. 3; Reardon Decl. 4; Trial Tr. 128–29, 538–39, 593–94, 675; July 28 Trial Tr. 10-11; PX005; PX007; PX024; DX056. Based on these communications, Nuance expected IBM to develop DeepQA into code that could be used to build commercial products, i.e., a single common core codebase that would be shared with Nuance regardless of which IBM group had developed the code. Ricci Decl. 6–7; Trial Tr. 493–94, 496–97; July 28 Trial Tr. 102; McCann Dep. 95–97, 101; PX11.079.
6. IBM was interested in licensing the DeepQA technology created by IBM Research to generate revenue to fund IBM Research, as it had done with prior technologies. Kelly Decl. 4–5; Trial Tr. 370. While the director of IBM Research had the authority to license technology outside of that group, he did not authorize the officers negotiating the DeepQA deal to license any updates created outside of IBM Research. Kelly Decl. 6–7; Trial Tr. 407; July 28 Trial Tr. 9; JX009. And Nuance was aware that IBM Research did not license technology held outside of IBM Research—including IBM Software code already contained in DeepQA—without the permission of the IBM group that held the technology being licensed. Reardon Decl. 5; King Decl. 2–3, ECF No. 216-4; Trial Tr. 165; Cox Dep. 104–05; DX10; DX110; DX122; DX123.
7. A provision entitling it to all IBM updates to DeepQA was integral to Nuance obtaining the commercially viable technology it sought because the DeepQA technology prepared for Jeopardy! was not ready for commercial applications. Bloom Decl. 5; Trial Tr. 87–88, 117–18, 225, 363, 377–78. And before entering the agreement, Nuance understood that its investment in DeepQA was risky and that the potential business applications of the technology were speculative. Trial Tr. 49–51, 56–58, 469; DX112. Accordingly, Nuance believed it needed assistance from IBM, in the form of updates to make DeepQA commercially usable, to successfully commercialize DeepQA. Petro Decl. 3; Trial Tr. 44–45.
8. When IBM's first draft of the proposed licensing agreement did not include a provision for updates to DeepQA, Nuance viewed it as a nonstarter. Bloom Decl. 6; Trial Tr. 185; JX004. Nuance insisted on including an updates provision—telling IBM that Nuance required the same updates provision as in a prior deal between the companies. Bloom Decl. 6; McCann Dep. 129–31; JX006.003. However, in that prior deal, Nuance had signed a separate agreement with IBM Software to obtain IBM Software-developed code. King Decl. 3; Trial Tr. 194–95; DX48.
9. In negotiating the DeepQA deal, Nuance expected the DeepQA technology to be transferred by IBM among its various groups, and Nuance wanted access to all updates to DeepQA, regardless of where they were developed within IBM. Ricci Decl. 6–7; Trial Tr. 166. And Nuance communicated to IBM that it wanted the updates provision to be as broad and long as possible, and that it expected updates to include new code. Cox Dep. 58–59; PX19.002.
10. That said, the parties understood the license did not include IBM-developed products; for the development of products, the parties entered a separate collaboration agreement, which Nuance ultimately terminated. King Decl. 5; McQueeney Decl. 2, 7; Trial Tr. 45, 200; McCann Dep. 91–96; DX19. And the parties understood the license to extend only to domain independent code—i.e., code that is generally applicable to all domains—as opposed to domain dependent code—e.g., code specific to the healthcare domain. Brown Decl. 4; Trial Tr. 71, 304.
11. The parties did not explicitly discuss which IBM groups would provide updates under the DeepQA deal: they did not specify Nuance would only be entitled to updates developed by IBM Research and not updates developed by IBM Software, but they also did not specify that Nuance would be entitled to updates developed by any IBM group. Ricci Decl. 8; Trial Tr. 39, 237, 553–54. And the parties did not discuss the possibility of IBM "forking" the DeepQA code—splitting the code into two separate components held in different IBM groups. July 28 Trial Tr. 11. Instead, IBM continued to reiterate its general commitment to investing in DeepQA to Nuance's benefit, including in a last-minute meeting between Nuance and IBM executives before the agreement was signed. Ricci Decl. 5–6; Trial Tr. 454, 493–94.
2. The Writing
12. On September 30, 2010, the parties signed the SLA, giving Nuance a perpetual license to the DeepQA technology and ten years of updates. JX001. The SLA is valid and binding, and it is governed by New York law. JX001.19; Joint Pretrial Order 4, ECF No. 155. It provides for a $25 million liability cap, regardless of the theory of liability, and precludes liability for any consequential damages. JX001.15–16. The agreement also includes an integration clause, providing that "[n]either party relies on any promises, inducements or representations made by the other, or expectations of more business dealings, except as expressly provided in this Agreement." JX001.18. And it provides for a two-year statute of limitations: "Neither party may bring an action arising out of this Agreement, regardless of form, more than two years after the cause of action has accrued and the party obtained knowledge thereof." JX001.18.
13. The SLA is an agreement between Nuance and IBM, "through its IBM Research Group." JX001.1. It provides that "all references to ‘IBM’ mean IBM Corporation, unless otherwise expressly limited to a division or group of IBM Corporation herein." JX001.19. Under the SLA, "IBM hereby grants to Nuance and its Subsidiaries a ... license to ... the Licensed IBM Background Software." JX001.2.
14. The SLA includes an updates provision by defining the licensed code to include updates for ten years. JX001.253–54. Specifically, it provides as follows:
"Licensed IBM Background Software" means (a) all Software that exists as of the Effective Date in all available formats (including Source Code and Object Code) that is owned by, or that has been developed or licensed by the IBM Research Group, including Tools, and that is listed on Exhibit A, including any modifications, updates, upgrades, error corrections, bug fixes, diagnostic and/or testing tools, that are JDBC compliant, and other changes, if available ("Modifications"), and if such Modifications are not contractually prohibited under a
Third Party agreement, and such, Modifications are available, will be timely provided to Nuance; and where the Modifications continue to meet the scope contemplated in Article 2.1 regarding the licensing of Deep QA under this Agreement, as of the Effective Date and thereafter for a period of ten (10) years, and additional Software as agreed by the parties, provided to Nuance by IBM under the Agreement (collectively "Updates"); and (b) all Software Materials for such Software.
JX001.253–54. The SLA also says that "[i]f IBM provides [ ] any modifications, updates, upgrades, error corrections, bug fixes, diagnostic and/or testing tools and other changes to the Licensed IBM Background Software, IBM will update Exhibit B to include any additions or subtractions to the Open Source Software or the Third Party Code." JX001.3.
D. Performance
15. A few months after the agreement was signed, DeepQA won Jeopardy!, and IBM determined that it had potential commercial applications that IBM wanted to pursue. Kelly Decl. 4–5; Trial Tr. 258–59.
1. IBM Creates Updates
16. Without informing Nuance, IBM "forked" the DeepQA code in 2011, which until then was held exclusively by IBM Research, giving IBM Software a copy of the code for it to begin commercialization efforts. Brown Decl. 9; McQueeney Decl. 3; Trial Tr. 270, 554, 671, 692–93, 697, 700; July 28 Trial Tr. 9, 99. This essentially created two separate versions of the code: the IBM Research version and the IBM Software version. Trial Tr. 693; Boloker Dep. 73–74, ECF No. 119-18. IBM invested significant resources into developing updates for the IBM Software version, including rewriting a significant amount of the DeepQA code. High Decl. 3–4, ECF No. 216-6; Trial Tr. 259; July 28 Trial Tr. 104.
17. The updates developed for the IBM Software version of the DeepQA code made the code commercially usable and added commercial adaptations. High Decl. 3–4; July 28 Trial Tr. 104.
18. First, between August 2011 and June 2012, the code was made commercially usable with updates made by IBM Software's Emerging Technologies Group. High Decl. 3; Trial Tr. 696; July 28 Trial Tr. 99; Boloker Dep. 53–54, 75–76, 83–84, 247. These updates involved "blue-washing" the code—i.e., deleting code that is not suitable for commercial use and replacing it with code that can be used in a commercial product—and making other updates that made the DeepQA code commercially viable. High Decl. 3; Schnell Decl. 14–15; Trial Tr. 261–63, 696; July 28 Trial Tr. 100; Boloker Dep. 54–72, 83; Rhodin Dep. 25–26, 29–30, ECF No. 119-17.
19. Then, between June 2012 and late 2015, IBM Software's Watson Division and the Watson Group developed updates that made the commercially usable DeepQA code applicable to business. High Decl. 3–4; July 28 Trial Tr. 104–05; Rhodin Dep. 29–36, 177–78. Unlike the commercial suitability updates, these commercial applicability updates were not precursors to making DeepQA into a product; instead, these latter updates adapted the code for commercial applications, which turned DeepQA into a product known as Watson Core. High Decl. 4–5; Kelly Decl. 6 n.1; July 28 Trial Tr. 104–05, 653; Boloker Dep. 57–72, 83; Rhodin Dep. 25–26, 29–33, 177–78; Xiang Dep. 103, ECF No. 60-9.
In 2014, the Watson Division of IBM Software was reorganized into the Watson Group. July 28 Trial Tr. 95.
20. Over several years, beginning in August 2014, IBM released two sets of Watson-branded products developed by IBM Software and Watson Group: one set that contains functionality related to DeepQA and another set that shares some code with DeepQA. High Decl. 4–7; July 28 Trial Tr. 106, 112; Liao Dep. 25, 110, 145, 175, 187, ECF No. 33-2. First, IBM Software developed products that perform some functions related to DeepQA's functions—e.g., the Retrieve and Rank product indexes and ranks documents in response to natural language questions, paralleling the search and answer-scoring stages of DeepQA. High Decl. 4–5; Schnell Decl. 37–39. Second, IBM Software developed products that use code that overlaps with DeepQA code—e.g., Watson for Oncology, which assists doctors in assessing treatment options for oncology patients, contains or uses some DeepQA code. Eggebraaten Decl. 2–5, ECF No. 216-8; Schnell Decl. 26–28.
IBM branded several DeepQA-related products as "Watson," in honor of IBM's founder Thomas J. Watson. See Mike Hale, Actors and Their Roles for $300, HAL? HAL!, N.Y. Times (Feb. 8, 2011), https://www.nytimes.com/2011/02/09/arts/television/09nova.html.
2. IBM Withholds Updates
21. But IBM did not give Nuance the Emerging Technologies Group updates, the Watson Division updates, or any of the Watson products. Trial Tr. 347, 426, 697; Boloker Dep. 72, 75. Instead, it withheld most DeepQA updates created outside of IBM Research from Nuance by not implementing a single common code base and thus preventing IBM Research from obtaining IBM Software's updates to DeepQA. Brown Decl. 5–6; Schnell Decl. 14; Trial Tr. 723–24.
22. To ensure that Nuance would not receive IBM Software's work on DeepQA, IBM implemented a firewall between IBM Research and IBM Software in 2011. Kelly Decl. 8; McQueeney Decl. 4; Trial Tr. 270–71, 273–74; July 28 Trial Tr. 74, 76-78; PX079.2. The firewall was one-sided, preventing IBM Research from obtaining IBM Software's updates to DeepQA, but allowing IBM Software to obtain IBM Research's updates. Brown Decl. 9–10; Trial Tr. 280. The implementation of a firewall was unusual because IBM Research ordinarily collaborated with IBM Software in the commercialization of IBM Research code, and the firewall was, accordingly, also counterproductive. McQueeney Decl. 4; Trial Tr. 272, 282, 705–06.
23. Because of the inefficiencies created by the firewall, IBM sought to amend the SLA. King Decl. 6; McQueeney Decl. 4. IBM told Nuance the amendment was meant to clarify the agreement, and it maintains that the amendment reflected what had always been IBM's intent, not a modification to the original agreement. Kelly Decl. 8–9; King Decl. 7; July 28 Trial Tr. 89; JX017.1. IBM's proposed amendment explicitly provided that Nuance would only be entitled to updates "that are written by one or more IBM Research Group employees or acquired under a contract entered into by IBM Research Group." JX017.2. And, in addition to a definition of IBM Research, it further provided as follows:
For avoidance of doubt, Software (including modifications, updates, upgrades, error corrections, bug fixes, diagnostic and/or testing tools) written by IBM employees that are not IBM Research Group employees shall not be considered "Licensed IBM Background Software" even if such non-IBM Research
Group employees consult with IBM Research Group employees in the course of their work on such Software. A "contract entered into by IBM Research Group" means a contract with a Third Party signed or authorized by IBM Research Group management to obtain code on behalf of IBM Research Group.
JX017.2.
24. Nuance rejected the amendment. Ricci Decl. 9; Joint Pretrial Order 4. Although IBM insists it did not understand Nuance to have rejected the amendment because it believed the amendment altered the agreement, Nuance told IBM it rejected the amendment because it preferred the contract as originally written. July 28 Trial Tr. 67–68, 89; JX019. And while IBM maintains that the amendment did not substantively change the agreement, it is not common for IBM to request non-substantive amendments to contracts. Kelly Decl. 8–9; King Decl. 7; Trial Tr. 287.
25. After Nuance rejected the amendment, IBM responded by investing few resources into developing updates for the IBM Research version of DeepQA, i.e., the version it shared with Nuance. Schnell Decl. 11; July 28 Trial Tr. 70–71; Boloker Dep. 53; JX019. IBM divested from the IBM Research version of the DeepQA code by moving several IBM Research engineers to IBM Software—and about 130 to Watson Group when it was formed—to work on the IBM Software version of the code. Trial Tr. 729–30; July 28 Trial Tr. 14, 95-96. And it further extracted intellectual capital from IBM Research by providing DeepQA tutorials to IBM Software and embedding IBM Software engineers in IBM Research for three months to learn about DeepQA. Brown Decl. 9; Trial Tr. 713, 728; July 28 Trial Tr. 134-35.
26. To be sure, IBM did give Nuance some updates developed outside of IBM Research—i.e., domain independent updates developed by the IBM Research team that had been transferred to the IBM Watson Group. Brown Decl. 5–6. Nonetheless, the hamstringing of IBM Research's efforts on DeepQA together with the withholding of IBM Software's updates to DeepQA from Nuance resulted in Nuance not receiving the commercially viable "blue-washed" DeepQA code. Petro Decl. 5; Ricci Decl. 9; Trial Tr. 131; July 28 Trial Tr. 70-71, 104.
27. Additionally, IBM also withheld some of the updates the pared-down IBM Research created after the divestment. Brown Decl. 7–8; Hicks Rebuttal Decl. 42–43, ECF No. 216-10; Schnell Decl. 18–20. IBM split the IBM Research version of the DeepQA code into two copies in 2012, storing them in separate folders to keep the updates that would be provided to Nuance separate from the updates withheld from Nuance. Brown Decl. 6; Schnell Decl. 18–20; Trial Tr. 762–65; PX070.3. It moved updates back and forth between these folders based on internal reconsiderations of what was an update under the SLA. Brown Decl. 8; Schnell Decl. 18–20. And IBM implemented an internal review process to ensure only the updates IBM believed counted as updates under the SLA were added to the folder that was shared with Nuance. Brown Decl. 7–8.
28. Some IBM employees charged with providing updates to Nuance did not understand what constituted updates under the SLA, and others understood enhancements and fixes to constitute the full extent of the SLA's definition of updates. Brown Decl. 8; Devarakonda Dep. 76, 85, 105. At least once, the decision to provide an update to Nuance was arbitrary, based only on whether IBM wanted to give an update to Nuance. PX89. Ultimately, IBM did not provide Nuance with some IBM Research updates that were related to DeepQA and improved the functionality of DeepQA. Brown Decl. 7–8; Schnell Decl. 19–20; Trial Tr. 762–65.
29. Nuance did not develop a commercial product from the DeepQA code. Ricci Decl. 9. Although Nuance invested resources into commercializing DeepQA by assigning employees and hiring consultants to work on the project, Nuance did not create a large DeepQA team internally, and some Nuance employees complained about a lack of effort to commercialize the code. Ricci Decl. 9; Trial Tr. 86–87, 90, 101; Fanty Dep. 68–69, 121–25; McCann Dep. 232–33; Stubley Dep. 75, 102, 189, ECF No. 131-17; DX124. Nuance's investments did not produce a commercial product from the DeepQA code. Ricci Decl. 9; Trial Tr. 117, 508.
3. The Parties’ Communications
30. At the same time as IBM withheld updates from Nuance, many communications related to DeepQA took place between the parties and internally among the parties’ employees.
31. Beginning in 2011, IBM communicated to Nuance that IBM Software was creating updates to DeepQA. McQueeney Decl. 5; DX71; DX88. Due to these communications, Nuance was aware that IBM Software was creating updates to DeepQA. Trial Tr. 92–93; McCann Dep. 189; Fanty Dep. 108; DX87; DX138.
32. IBM also communicated to Nuance during this time that Nuance would not be receiving IBM Software's DeepQA work. Kelly Decl. 10; McQueeney Decl. 5, 7; Fanty Dep. 43, 83; McCann Dep. 166–67; DX103. And Nuance was aware that IBM was not providing it with work done by IBM Software. DX84; DX103.4; DX126; DX138. Indeed, IBM told Nuance that a firewall had been erected between IBM Research and IBM Software to prevent Nuance from obtaining IBM Software's work product. Kelly Decl. 10; Fanty Dep. 41–43, 83–84, 107–08, 152–53; McCann Dep. 199–200, 214; Rhodin Dep. 73–76; DX103.4; DX138.
33. With this knowledge, Nuance sought assurances from IBM that it would not breach the agreement, but it only sought non-specific assurances and received non-specific assurances. Ricci Decl. 10; Trial Tr. 239–40, 299–300; McCann Dep. 184, 235. Nuance did not have access to IBM's DeepQA code to verify which updates were not being provided and whether they constituted updates under the SLA. Bloom Decl. 10; Petro Decl. 5; McCann Dep. 184. So, on several occasions Nuance asked IBM whether IBM was delivering all DeepQA updates, without specifying that "all updates" included updates created outside of IBM Research, and IBM affirmed that it was delivering all updates under the SLA, without specifying it was withholding all updates created outside of IBM Research. Trial Tr. 298—302, 397, 766–67; McCann Dep. 184, 196–99, 235, 238–40; DX039; PX084.
34. On April 13, 2015, IBM told Nuance explicitly that the IBM Software version of the DeepQA code was compatible with products developed by IBM. IBM further told Nuance that Nuance's version of the code was not compatible with these products. Ricci Decl. 10; Trial Tr. 429–32, 551; Sejnoha Dep. 96–97, ECF No. 60-3; PX105. Nuance filed suit on June 30, 2016. Compl. 1.
E. Updates and Products
35. The term "product" is not defined in the SLA. The common and ordinary understanding of "product" is "something (such as a service) that is marketed or sold as a commodity." See Product, Merriam-Webster's Collegiate Dictionary, https://unabridged.merriam-webster.com/collegiate/product (last visited June 14, 2021). On the other hand, the SLA does provide a definition of "updates." See supra Section III.C.2. IBM's expert witness and Nuance's expert witness each provided one-sentence opinions about the definitions of the key terms in the SLA (e.g., "updates" and "modifications"), without fully agreeing on the definitions of these terms, see Hicks Decl. 5-7, and the parties failed to sufficiently corroborate their respective definitions of these terms with testimony from fact witnesses.
36. There are three categories of code that Nuance claims it is entitled to and that are at issue in this case.
37. First, there are the updates, created by IBM Software, that culminated in the Watson Core product. High Decl. 3–7; Kelly Decl. 6 n.1; Trial Tr. 653. Between August 2011 and June 2012, IBM developed updates that made DeepQA commercially usable by removing dead code, removing open-source code, replacing hard code, adding tooling, and making other changes that made the code work outside of the Jeopardy! context. High Decl. 3; Schnell Decl. 15; Trial Tr. 261–64, 696; July 28 Trial Tr. 100; Boloker Dep. 54–72, 83; Rhodin Dep. 25–26, 29–30. Then, from June 2012 to late 2015, IBM developed updates that made DeepQA adaptable to business, adding code that enabled multi-word answers and added dialogue capabilities, as well as other functions that made the code into a commercialized product. High Decl. 4–5; July 28 Trial Tr. 104–05; Boloker Dep. 57–72, 83; Rhodin Dep. 25–26, 29–33, 177–78. The question-answering technology that resulted from all these updates is referred to as Watson Core. Kelly Decl. 6 n.1; Trial Tr. 653; Xiang Dep. 103.
38. Second, there are the updates, created by IBM Research, that IBM determined were not updates under the SLA. Brown Decl. 8; Hicks Rebuttal Decl. 42–43; Schnell Decl. 18–20. These updates were on topics related to DeepQA or shared functionality with DeepQA and would be desired for DeepQA. Schnell Decl. 19–21.
39. Third, there are the following products, which either (1) have similar functions with DeepQA, (2) improve on the functionality of DeepQA, (3) share code with DeepQA, or (4) use code from DeepQA. See Hicks Rebuttal Decl. 16, 18, 20, 25; Schnell Decl. 22-29, 34, 36-37, 39-41, 43; Morello Dep. 31, ECF No. 33-3; Liao Dep. 210; Trial Tr. 740; July 28 Trial Tr. 108, 111, 125, 143, 147-50; July 30 Trial Tr. 106, 194-95.
• Document Conversion converts electronic documents into formats that can be read by question-answering technology, allowing users to upload their own corpus of information from which the technology will provide answers to the users’ questions. July 28 Trial Tr. 111; July 30 Trial Tr. 195; Liao Dep. 25–27.
• Watson Knowledge Studio allows subject matter experts to build custom models that understand the linguistic subtleties of specific domains or industries. Schnell Decl. 41–43; July 30 Trial Tr. 200; Liao Dep. 110.
• Watson Explorer collects and compiles data from different sources and returns documents as answers to a user's natural language question. Schnell Decl. 40–41; July 30 Trial Tr. 200; Liao Dep. 179–80, 210.
• Natural Language Understanding provides APIs that analyze text and extract semantic features from
API, or application program interface, is software that allows the user to get access to the function or service that the API represents. July 28 Trial Tr. 122.
the text, such as concepts, categories, and sentiment. Schnell Decl. 36–37; July 30 Trial Tr. 199; Liao Dep. 146–47.
• Retrieve and Rank retrieves relevant documents in response to a natural language question by indexing a corpus of documents and presenting the user with the documents most relevant to the question. Schnell Decl. 37–39; July 30 Trial Tr. 197–98; Morello Dep. 37–38.
• Discovery converts electronic documents into a readable format and ranks and returns relevant documents in response to natural language questions; it replaces the Document Conversion and Retrieve and Rank products and enhances their functions with machine learning. Schnell Decl. 34–36; July 30 Trial Tr. 198–99; Liao Dep. 65–66.
• Tone Analyzer allows natural language technology to detect emotion, social tendencies, and writing style in text. High Decl. 6–7; Schnell Decl. 39–40; July 30 Trial Tr. 199–200.
• Natural Language Classifier allows natural language technology to understand the user's question when the same question is asked in different ways, by classifying the intent of expressions. Schnell Decl. 28–31; Xiang Dep. 53.
• Watson Engagement Advisor directs customers to products or services based on their expressions of their needs, using the DeepQA concept with added capabilities for descriptive answers and retrieval of excerpts from documents. High Decl. 4; July 28 Trial Tr. 106–08.
• Watson Discovery Advisor assists users with research by ranking the evidence supporting an answer in response to a natural language question. High Decl. 5; July 28 Trial Tr. 112.
• Watson for Oncology assists doctors in assessing treatment options for oncology patients by returning ranked treatment options in response to patient clinical information. Eggebraaten Decl. 2; July 28 Trial Tr. 119.
IV. CONCLUSIONS OF LAW AND DISCUSSION
The Court has subject-matter jurisdiction, venue is proper, and New York law applies to the case.
A breach of contract claim under New York contract law requires proof of "(i) the formation of a contract between the parties; (ii) performance by the plaintiff; (iii) failure of defendant to perform; and (iv) damages." Orlander v. Staples, Inc., 802 F.3d 289, 294 (2d Cir. 2015) (quoting Johnson v. Nextel Commc'ns, Inc., 660 F.3d 131, 142 (2d Cir. 2011) ); see also Stonehill Cap. Mgmt. LLC v. Bank of the W., 28 N.Y.3d 439, 45 N.Y.S.3d 864, 68 N.E.3d 683, 688 (2016). It is undisputed that a contract was formed and that Nuance performed. Joint Pretrial Order 4. Thus, before the Court are issues of contract interpretation and breach of contract. Because IBM raises a statute of limitations defense, this issue is also before the Court.
Although the parties also raise the issue of what remedies are appropriate, the Court need not address this issue because Nuance's claims are barred by the statute of limitations, as further discussed below.
A. Breach of Contract
1. Contract Interpretation
General principles of contract interpretation apply under New York law. When "interpreting a contract, the intent of the parties governs, ... [a] contract should be construed so as to give full meaning and effect to all of its provisions, [and] words and phrases are given their plain meaning." PaineWebber Inc. v. Bybyk, 81 F.3d 1193, 1199 (2d Cir. 1996) (first alteration in original) (quoting Am. Express Bank Ltd. v. Uniroyal, Inc., 164 A.D.2d 275, 562 N.Y.S.2d 613, 614 (1990) ) (first citing Tigue v. Com. Life Ins. Co., 219 A.D.2d 820, 631 N.Y.S.2d 974, 975 (1995) ; and then citing Heller v. Pope, 250 N.Y. 132, 164 N.E. 881 (1928) ). Further, a contract is "interpreted in the light of all the circumstances, and if the principal purpose of the parties is ascertainable it is given great weight." Rockland Exposition, Inc. v. All. of Auto. Serv. Providers of N.J., 706 F. Supp. 2d 350, 359 (S.D.N.Y. 2009) (quoting Restatement (Second) of Contracts § 202(1) (Am. L. Inst. 1981), which New York courts rely on in breach of contract cases, see, e.g., Maas v. Cornell Univ., 94 N.Y.2d 87, 699 N.Y.S.2d 716, 721 N.E.2d 966 (1999) ); see also Hooper Assocs., Ltd. v. AGS Computs., Inc., 74 N.Y.2d 487, 549 N.Y.S.2d 365, 548 N.E.2d 903, 905 (1989) ("Words in a contract are to be construed to achieve the apparent purpose of the parties."). When the contract is ambiguous on its face as a matter of law, the meaning of the ambiguous provisions must be determined by a factfinder, and the factfinder can consider extrinsic evidence.
See JA Apparel Corp. v. Abboud, 568 F.3d 390, 397 (2d Cir. 2009) ("However, where the contract language creates ambiguity, extrinsic evidence as to the parties’ intent may properly be considered." (first citing Seiden Assocs., Inc. v. ANC Holdings, Inc., 959 F.2d 425, 426, 429 (2d Cir. 1992) ; then citing In re Consol. Mut. Ins. Co., 77 N.Y.2d 144, 565 N.Y.S.2d 434, 566 N.E.2d 633, 635 (1990) ; and then citing 67 Wall St. Co. v. Franklin Nat'l Bank, 37 N.Y.2d 245, 371 N.Y.S.2d 915, 333 N.E.2d 184, 186 (1975) )); 3Com Corp. v. Banco do Brasil, S.A., 171 F.3d 739, 746 (2d Cir. 1999) ("It is often stated that the existence of ambiguity in contractual language is a ‘question of law’ for the court, and the resolution of any such ambiguity is a ‘question of fact’ for the jury." (quoting Bouzo v. Citibank, N.A., 96 F.3d 51, 58 (2d Cir. 1996) )); Blue Jeans U.S.A., Inc. v. Basciano, 286 A.D.2d 274, 729 N.Y.S.2d 703, 705 (2001) ("[W]hen a contract term is ambiguous, parol evidence may be considered ‘to elucidate the disputed portions of the parties’ agreement.’ " (citation omitted)).
Extrinsic evidence includes (1) "the acts and circumstances surrounding execution of the ambiguous term," (2) "conversations, negotiations and agreements made prior to or contemporaneous with the execution of a written [agreement]," and (3) "the parties’ course of conduct throughout the life of the contract." GE Funding Cap. Mkt. Servs., Inc. v. Neb. Inv. Fin. Auth., No. 15-1069, 2017 WL 2880555, at *4 (S.D.N.Y. July 6, 2017) (alteration in original) (first quoting Roberts v. Consol. Rail Corp., 893 F.2d 21, 24 (2d Cir. 1989) ; then quoting Mooney v. AXA Advisors, LLC, 19 F. Supp. 3d 486, 506 (S.D.N.Y. 2014) (quoting 67 Wall St. Co. v. Franklin Nat'l Bank, 37 N.Y.2d 245, 371 N.Y.S.2d 915, 333 N.E.2d 184, 186 (1975) ); and then quoting Hoyt v. Andreucci, 433 F.3d 320, 332 (2d Cir. 2006) ), aff'd, 767 F. App'x 110 (2d Cir. 2019).
Trade usage may be considered as extrinsic evidence if "the other party was actually aware of the trade usage, or ... the usage was so notorious in the industry that a person of ordinary prudence in the exercise of reasonable care would be aware of it." GE Funding, 2017 WL 2880555, at *4 (quoting Last Time Beverage Corp. v. F & V Distrib. Co., 98 A.D.3d 947, 951 N.Y.S.2d 77, 81–82 (2012) ). Neither party has provided sufficient evidence that trade usage related to updates, modifications, or any other term at issue was understood by the other party or was so notorious that a person of ordinary prudence would be aware of it. See supra Section III.E. Therefore, both parties failed to meet the preponderance of the evidence standard and the Court does not consider evidence of trade usage in defining these terms.
The text of the provision which is the subject of the dispute is as follows:
"Licensed IBM Background Software" means (a) all Software that exists as of the Effective Date in all available formats (Including Source Code and Object Code) that is owned by, or that has been developed or licensed by the IBM Research Group, including Tools, and that is listed on Exhibit A, including any modification, updates, upgrades, error corrections, bug fixes, diagnostic and/or testing tools, that are JDBC compliant, and other changes, if available ("Modifications"), and if such Modifications are not contractually prohibited under a Third Party agreement, and such Modifications are available, will be timely provided to Nuance; and where the Modifications continue to meet the scope contemplated in Article 2.1 regarding the licensing of Deep QA under this Agreement, as of the Effective Date and thereafter for a period of ten (10) years, and additional Software as agreed by the parties, provided to Nuance by IBM under the Agreement (collectively "Updates") ....
JX001.253–54.
The parties present opposing views about whether this language entitles Nuance to updates from only IBM Research or from all of IBM. Nuance argues that this provision is separated into two parts by the semicolon, the first part defining modifications and the second part providing that for a period of ten years, IBM will give Nuance these modifications, i.e., updates. Summ. J. Op. 21. IBM contends that the provision is separated into two parts by the second to last comma, the first part defining IBM Research modifications to which Nuance is entitled under the agreement and the second part providing that IBM will provide Nuance additional code pursuant to any separate subsequent agreements. Summ. J. Op. 20. The SLA is ambiguous on its face, and a genuine dispute of material fact exists about the parties’ intent after considering the extrinsic evidence. Summ. J. Op. 23, 27.
The Court finds by the preponderance of the evidence that this provision means Nuance is entitled to updates in the sense of code that is added to DeepQA that made DeepQA commercially usable, but not to products that are similar to or derived from DeepQA. Thus, finished products, such as Watson Core, Tone Analyzer, and Watson for Oncology, see supra Sections III.D.1, III.E, were not updates under the agreement, but non-product updates, such as the blue-washed code and other similar updates, see supra Sections III.D.1-2, III.E, could be updates under the agreement if they made DeepQA commercially usable. See supra Section III.E.
The pre-contractual discussions and negotiations show the parties agreed to give Nuance a license to DeepQA code that was usable in commercial products. The Court finds this was the parties’ intent based on Nuance's expressions of interest in commercially viable code, not code with only game show capabilities, and IBM's representations that it would invest in DeepQA to make it commercially viable. That said, the parties did not intend to give Nuance any products developed by IBM, even if the product borrowed a significant amount of code from DeepQA (e.g., Watson Discovery Advisor) or was a productized version of the DeepQA code (e.g., Watson Core). To that end, the parties were explicit that Nuance would only be entitled to domain independent code and the parties were negotiating a separate agreement for products. While the parties did not intend to give Nuance access to all the natural language technology developed by IBM, their primary purpose was to give Nuance access to any updates to DeepQA that would facilitate Nuance's creating commercially applicable products directly from the DeepQA code (e.g., the blue-washed code).
This mutual understanding about the agreement—that Nuance would receive commercially applicable code—did not limit Nuance to receiving updates created by IBM Research. Nuance communicated to IBM that it wanted the updates to be broad, and the parties never explicitly said the updates would be limited to those developed by IBM Research. IBM relies on Nuance's general understanding that IBM's groups operate as separate units. Based on that understanding, however, Nuance sought to ensure that the updates would come not only from IBM Research, but from all of IBM's units, by referring to IBM and not only IBM Research in the applicable provisions.
The parties’ conduct of performance confirms their mutual intent was to give Nuance a license to DeepQA code that was usable in commercial products and not a license to finished products developed from DeepQA. IBM forked the code and implemented a firewall to ensure the products it developed from the DeepQA code, such as Watson Discovery Advisor and the Watson Core, see supra Sections III.D, III.E, were not provided to Nuance. And Nuance accepted this firewall and understood it served the purpose of protecting IBM's products, which were ordinarily made by IBM Software. But IBM did not explicitly tell Nuance that it would not receive any non-product updates, such as the blue-washed code, see supra Section III.E, created outside of IBM Research, and IBM in fact continued to give Nuance non-product updates, including some created outside of IBM Research.
Further, the proposed amendment shows Nuance and IBM understood the agreement to not limit updates to those created by IBM Research. And IBM's response after Nuance rejected the amendment—i.e., IBM's divestment from IBM Research and diversion of investments to IBM Software—does not contradict this understanding, but instead shows IBM understood Nuance was entitled to updates from any unit within IBM (i.e., not just IBM Research), but not products, which was IBM Software's focus. Thus, by investing in products, IBM sought to prevent Nuance from obtaining IBM's work on natural language technology when the parties’ relationship soured over the amendment dispute.
2. Breach
Because the SLA entitles Nuance to non-product updates, made anywhere in IBM, that make DeepQA commercially usable, IBM breached the agreement by withholding the updates made by IBM Software that prepared DeepQA for commercial applications. This includes only the updates developed by IBM Software between August 2011 and June 2012—i.e., those developed during the blue-washing of the DeepQA code—because these are the only updates Nuance has shown were not products and made DeepQA commercially viable. Nuance showed by the preponderance of the evidence that the blue-washed code was a software architecture that was not marketable as a product, unlike the Watson Core, and that the blue-washed code made the DeepQA code commercially viable by making it applicable beyond Jeopardy! to business solutions. See supra Sections III.D.1, III.E. Nuance's obtaining the blue-washed code would have given it the benefit it sought from the bargain—i.e., commercially usable code—and thus the withholding of the blue-washed code frustrated the purpose of the agreement.
See Orlander v. Staples, Inc., 802 F.3d 289, 298 (2d Cir. 2015) ("In any event, the question of the materiality of a breach ‘is usually a question of fact and should be decided as a matter of law only where the inferences are certain.’ " (citation omitted)).
Nuance failed to show that any software other than the blue-washed code made DeepQA commercially usable and was not a product. The evidence does not establish that the products IBM developed, such as Watson Core and Watson for Oncology, were not marketable products—and thus these were not updates under the agreement. To the extent IBM did not give Nuance other non-product updates created in IBM Research, see supra Sections III.D.2, III.E, Nuance did not produce evidence that these updates made DeepQA commercially viable—thus, these were also not updates under the agreement. Therefore, IBM's failure to give Nuance the blue-washed DeepQA code was a material breach of the SLA.
Additionally, even if the terms of the SLA did not expressly entitle Nuance to updates developed by IBM Software, the preponderance of the evidence shows that IBM's conduct constitutes a breach of contract through a breach of the implied covenant of good faith because its conduct effectively deprived Nuance of the benefit of the bargain. Under New York law, all contracts have an implied covenant of good faith and fair dealing, and a breach of that covenant is "a breach of the underlying contract." Fasolino Foods Co. v. Banca Nazionale del Lavoro, 961 F.2d 1052, 1056 (2d Cir. 1992) (quoting Geler v. Nat'l Westminster Bank USA, 770 F. Supp. 210, 215 (S.D.N.Y. 1991) (citing Murphy v. Am. Home Prods. Corp., 58 N.Y.2d 293, 461 N.Y.S.2d 232, 448 N.E.2d 86, 91 (1983) )). To be sure, the implied covenant "cannot be construed so broadly as effectively to nullify other express terms of a contract, or to create independent contractual rights." Fesseha v. TD Waterhouse Inv. Servs., 305 A.D.2d 268, 761 N.Y.S.2d 22, 23 (2003) (citations omitted). While the implied covenant does not prevent a party from acting in its own interest, it prevents a party from breaking an implied promise that "is so interwoven" with an express promise in the contract that the conduct destroys the ability of the nonbreaching party to receive the benefit of the express promise. Thyroff v. Nationwide Mut. Ins. Co., 460 F.3d 400, 407–08 (2d Cir. 2006) (citing M/A-COM Sec. Corp. v. Galesi, 904 F.2d 134, 136 (2d Cir. 1990) (quoting Havel v. Kelsey-Hayes Co., 83 A.D.2d 380, 445 N.Y.S.2d 333, 336 (1981) )).
Although the separate claim for breach of the implied covenant of good faith and fair dealing was dismissed at summary judgment, Nuance may still recover under its breach of contract claim for conduct that constitutes a breach of the implied covenant. A claim for breach of the implied covenant of good faith that is based on the same facts and that seeks the same remedies as a claim for breach of contract will be dismissed as duplicative. Arcadia Biosciences, Inc. v. Vilmorin & Cie, 356 F. Supp. 3d 379, 400 (S.D.N.Y. 2019) (quoting Houbigant, Inc. v. ACB Mercantile (In re Houbigant, Inc.), 914 F. Supp. 964, 989 (S.D.N.Y. 1995) (citing R.H. Damon & Co. v. Softkey Software Prods., Inc., 811 F. Supp. 986, 992 (S.D.N.Y. 1993) (citing Tesoro Petroleum Corp. v. Holborn Oil Co., 108 A.D.2d 607, 484 N.Y.S.2d 834, 835 (1985) ))). But where the meaning of a contract is in doubt, courts allow a breach of contract claim and a breach of the implied covenant of good faith claim to be pleaded in the alternative. Hard Rock Cafe Int'l, (USA), Inc. v. Hard Rock Hotel Holdings, LLC, 808 F. Supp. 2d 552, 567 (S.D.N.Y. 2011). And there may be a breach of an express provision through conduct that constitutes a breach of the implied covenant. See Ret. Bd. of Policemen's Annuity & Benefit Fund of City of Chi. v. Bank of N.Y. Mellon, No. 11-5459, 2014 WL 3858469, at *4 (S.D.N.Y. July 30, 2014) (noting that a claim for breach of the implied covenant is "inherent in every breach of contract claim" and that a "viable breach of contract claim [may be] based on [a breach of the] implied covenant" (citing ABN AMRO Bank, N.V. v. MBIA Inc., 17 N.Y.3d 208, 928 N.Y.S.2d 647, 952 N.E.2d 463, 474-75 (2011) )).
In this case, Judge Karas granted IBM's Motion for Summary Judgment pertaining to the implied covenant claim because it was duplicative and redundant of the breach of contract claim. See Summ. J. Op. 29-31. Thus, the Court will not consider the implied covenant claim as a separate claim, but rather as an alternative basis for liability under the breach of contract claim.
See also Reilly v. NatWest Mkts. Grp. Inc., 181 F.3d 253, 263 (2d Cir. 1999) (holding that claims under contract and quasi-contract may be pleaded in the alternative when the meaning of the contract is unclear (citing N.Y. C.P.L.R. 3002 (McKinney 1991) )).
See also Carvel Corp. v. Diversified Mgmt. Grp., Inc., 930 F.2d 228, 230 (2d Cir. 1991) (holding that it was reversible error for the trial court to not give the jury a good faith instruction on a breach of contract claim).
The Court finds by the preponderance of the evidence that IBM breached the implied covenant of good faith in the SLA by withholding non-product updates from Nuance that made the DeepQA code commercially usable (i.e., the blue-washed code). In response to Nuance not agreeing to amend the terms of the SLA, IBM moved knowledgeable engineers out of IBM Research and into IBM Software, effectively limiting IBM Research from creating updates to which Nuance would be entitled. Instead, IBM invested in IBM Software creating products (e.g., Watson Explorer) that Nuance would not be entitled to receive. In the process of developing these products, however, IBM created updates to DeepQA that made the code commercially viable without making it a product—specifically, in making the Watson Core product, IBM developed the blue-washed code updates. Thus, IBM created updates to DeepQA under the SLA when it created the blue-washed code updates that made DeepQA commercially usable, and it breached the agreement when it did not provide these updates to Nuance. IBM took these steps in bad faith for the purpose of depriving Nuance of the benefit of the bargain—DeepQA code that was commercially usable—in response to Nuance not acceding to IBM's request to amend the contract.
Nuance initially brought three counts: declaratory judgment, breach of contract, and breach of the implied covenant of good faith and fair dealing. Judge Karas dismissed the implied covenant claim as duplicative of the breach of contract claim, so this Court now only has two counts to consider: the declaratory judgment claim and the breach of contract claim (which includes the implied covenant claim). Nuance's declaratory judgment claim that IBM is required to deliver all updates to the Licensed IBM Background Software, regardless of where they are developed within IBM, is essentially a request for a declaratory judgment that IBM breached its contract with Nuance. Courts in New York apply the same statute of limitations to declaratory judgment claims that they do to the underlying claim for breach of contract. See, e.g., Gandhi v. Nayak, 148 A.D.2d 390, 539 N.Y.S.2d 335, 336 (1989) ("Since this is a declaratory judgment action, we find that the six-year Statute of Limitations, applicable to the plaintiff's breach of contract claim, which underlies the instant action, controls the time in which this action must be commenced." (citations omitted)); Meadowbrook Farms Homeowners Ass'n v. JZG Res., Inc., 105 A.D.3d 820, 963 N.Y.S.2d 300, 302 (2013) (holding that "[t]he same six-year statute of limitations governs the plaintiff's declaratory judgment causes of action." (citation omitted)). Thus, given the Court's decision that the breach of contract claim is barred by the statute of limitations, the declaratory judgment claim, which is subject to the same statute of limitations, is also barred.
Under New York law, there is a six-year statute of limitations period for breach of contract suits, but this limitations period can be shortened by contract, and "the shorter period controls as long as it is reasonable." Ajdler v. Province of Mendoza, 890 F.3d 95, 99 (2d Cir. 2018) (first citing N.Y. C.P.L.R. 201 (Consol. 2021); then citing John J. Kassner & Co. v. City of New York, 46 N.Y.2d 544, 415 N.Y.S.2d 785, 389 N.E.2d 99, 103 (1979) ; and then citing Executive Plaza, LLC v. Peerless Ins. Co., 22 N.Y.3d 511, 982 N.Y.S.2d 826, 5 N.E.3d 989, 991 (2014) ), aff'd, 768 F. App'x 78 (2d Cir. 2019). Ordinarily, the statute of limitations for a breach of contract claim begins running once the contract is breached, regardless of the injured party's knowledge. T&N PLC v. Fred S. James & Co., 29 F.3d 57, 59–60 (2d Cir. 1994) (citing Ely-Cruikshank Co. v. Bank of Montreal, 81 N.Y.2d 399, 599 N.Y.S.2d 501, 615 N.E.2d 985, 986 (1993) ); see also ABB Indus. Sys. v. Prime Tech., 120 F.3d 351, 360 (2d Cir. 1997) ("[I]n New York it is well settled that the statute of limitation for breach of contract begins to run from the day the contract was breached, not from the day the breach was discovered, or should have been discovered." (first citing Ely-Cruikshank Co., 599 N.Y.S.2d 501, 615 N.E.2d at 968-88; and then citing Nat'l Life Ins. Co. v. Frank B. Hall & Co. of N.Y., 67 N.Y.2d 1021, 503 N.Y.S.2d 318, 494 N.E.2d 449, 450 (1986) )). But while the limitations period ordinarily begins to run from the moment the cause of action accrues, clear language in the contract that sets the start of the limitation clock at another time will control. See Fabozzi v. Lexington Ins. Co., 601 F.3d 88, 91 (2d Cir. 2010) (citing Steen v. Niagara Fire Ins. Co., 89 N.Y. 315, 322-24 (1882) ).
Here, there is a reasonable and enforceable contractual limitations period of two years from when the plaintiff obtained actual knowledge of the alleged breach. Summ. J. Op. 10-11, ECF No. 150. Nuance filed suit on June 30, 2016. Nuance claims that it did not obtain actual knowledge until 2015, when IBM told Nuance that "there was a whole corpus of core technology that we didn't have access to, that there, in fact, was a separate [DeepQA] core." Trial Tr. 551. Thus, to prevail on the statute of limitations defense, IBM must show that before June 30, 2014, Nuance obtained actual knowledge of the alleged breach.
Actual knowledge may be proven by circumstantial evidence, and evidence of willful blindness can be considered in determining actual knowledge. See U.S. Bank, Nat'l Ass'n v. UBS Real Est. Sec. Inc., 205 F. Supp. 3d 386, 425 (S.D.N.Y. 2016) (quoting Woloszynowski v. N.Y. Cent. R.R. Co., 254 N.Y. 206, 172 N.E. 471, 472-73 (1930) ) (citing Reed v. Fed. Ins. Co., 123 A.D.2d 188, 510 N.Y.S.2d 618, 622 (1987), aff'd, 71 N.Y.2d 581, 528 N.Y.S.2d 355, 523 N.E.2d 480 (1988) ).
Given that the parties agree that actual knowledge is the applicable test here, this case is distinct from Meda AB v. 3M Co., 969 F. Supp. 2d 360 (S.D.N.Y. 2013). Meda was a breach of contract case in which the plaintiff argued that the defendant breached a representation in their contract that stated " ‘[t]o Seller's Knowledge, the Business is not in violation of any Law.’ ‘Seller's Knowledge’ is defined in the [contract] as ‘the actual knowledge of any individuals without inquiry listed in Section 1.01(e) of the Seller Disclosure Schedule,’ ...." Id. at 380 (first alteration in original). The court held that the plaintiff's attempt to impose a willful blindness theory of liability onto the representation was inconsistent with the definition of "Seller's Knowledge" in the contract, given that the contract only referred to what defendants actually knew "without inquiry." Id. Here, there is no such similar "without inquiry" language in the contract between Nuance and IBM.
1. Willful Blindness
a. The Willful Blindness Doctrine
The doctrine of willful blindness originated in the criminal context and can be useful in establishing the mens rea for a variety of crimes. See, e.g., United States v. Rivera, 944 F.2d 1563, 1573 (11th Cir. 1991) (treating "deliberate ignorance," a corollary of willful blindness, as a state of mind equally culpable as actual knowledge); United States v. Rothrock, 806 F.2d 318, 323 (1st Cir. 1986) ("The purpose of the willful blindness theory is to impose criminal liability on people who, recognizing the likelihood of wrongdoing, nonetheless consciously refuse to take basic investigatory steps."); United States v. One 1973 Rolls Royce, V.I.N. SRH-16266 by & Through Goodman, 43 F.3d 794, 808-09 (3d Cir. 1994) (recognizing that "the mainstream conception of willful blindness [is] a state of mind of much greater culpability than simple negligence or recklessness, and more akin to knowledge" (first citing Rivera, 944 F.2d at 1570 ; and then citing Rothrock, 806 F.2d at 323 )).
"Under New York law, where a party ‘had sufficient information to impose a duty upon it to make further inquiry ... its failure to do so constituted "willful blindness." ’ " U.S. Bank, 205 F. Supp. 3d at 425 (quoting Scher L. Firm, LLP v. DB Partners I, LLC, 97 A.D.3d 590, 948 N.Y.S.2d 335, 337 (2012) ). "[A] willfully blind defendant is one who takes deliberate actions to avoid confirming a high probability of wrongdoing and who can almost be said to have actually known the critical facts." Id. (alteration in original) (quoting Scher L. Firm, 948 N.Y.S.2d at 338 ) (citing Viacom Int'l, Inc. v. YouTube, Inc., 676 F.3d 19, 35 (2d Cir. 2012) ).
On the civil side, the doctrine has venerable roots in New York dating back to the late 1800s or early 1900s. The most applicable case is Reed v. Gannon, 50 N.Y. 345 (1872). In Reed, under a separation agreement between a husband and wife, the husband transferred certain property to the wife and agreed to pay off any existing mortgages within ninety days. Id. at 348. The husband had mortgaged the property to a creditor and failed to pay off the debt. See id. at 349. The creditor then instituted foreclosure proceedings on the property, and the wife filed an action to restrain the foreclosure, arguing that she was a bona fide transferee. See id. at 351. The trial court found in favor of the wife, and the appellate court affirmed. See id. at 352. The creditor appealed. See id. The New York Court of Appeals reversed and remanded for a new trial, holding that the wife had actual notice that mortgages existed based on the language in the separation agreement. See id. The court further held that the wife's failure to inquire as to the extent and description of the mortgages constituted willful blindness, so the wife was deemed to have notice as to the extent and description of the mortgages as well. See id. at 349-51 ("The omission to make the proper examination or inquiry is gross negligence, which prevents the purchaser from being regarded as bona fide .... If the natural and obvious inquiry was not even made of Reed, the vendor, it was one of the cases characterized in the authorities as willful blindness.").
The doctrine in New York is regularly applied in the civil context to cases concerning investment securities under New York's Uniform Commercial Code ("U.C.C.") section 8-105(a)(2), which states that a person has notice of an adverse claim if "the person is aware of facts sufficient to indicate that there is a significant probability that the adverse claim exists and deliberately avoids information that would establish the existence of the adverse claim." N.Y. U.C.C. Law § 8-105(a)(2) (Consol. 2021). The court in Scher Law Firm noted that the Second Circuit has described this provision as the "willful blindness" test. 948 N.Y.S.2d at 338 (quoting SEC v. Credit Bancorp, Ltd., 386 F.3d 438, 448 (2d Cir. 2004) ).
This Scher Law Firm formulation of the doctrine, see id., has been extended by New York state courts to cases falling outside of the U.C.C. Article 8 context. For example, in Wilmington Trust v. MC-Five Mile Commercial Mortgage Finance LLC, 171 A.D.3d 591, 99 N.Y.S.3d 11 (2019), the defendant breached a loan contract because a lockbox agreement required by the contract was not timely completed, which constituted an event of default under the contract. The defendant denied knowledge that the lockbox agreement was not timely completed. See id. at 12. However, the defendant was a required party to the lockbox agreement, so the court found that it must have known that it never signed the agreement. Id. The court held that the plaintiff was entitled to partial summary judgment on its cause of action seeking specific performance of the contract because "[a]t the very least, these facts demonstrate that defendant was willfully blind with respect to the status of the lockbox agreement, which is evidence that defendant had knowledge of that status." Id. (emphasis added) (citations omitted).
More recently, the doctrine was applied by another New York state court in a breach of contract case, Daimler Trust v. Progressive Casualty Insurance Co., No. 901931-19, 2020 N.Y. Misc. LEXIS 10959 (Sup. Ct. Nov. 5, 2020). In Daimler Trust, the plaintiff's insurance company, Progressive, wrongly identified the plaintiff as the owner of the vehicle that the plaintiff insured with Progressive. Id. at *3. The plaintiff argued that Progressive's failure to review any document other than the plaintiff's application for insurance, such as title or registration, amounted to willful blindness. Id. The court granted Progressive's motion for summary judgment, finding that Progressive was not willfully blind because the plaintiff did not allege or provide any proof that Progressive had actual knowledge of the plaintiff's lack of an ownership interest in the vehicle. Id. at *13.
This is unlike the present case, given that IBM alleges actual knowledge and has provided evidence to support this allegation.
Federal courts in New York have also applied the doctrine. U.S. Bank, National Association v. UBS Real Estate Securities Inc., 205 F. Supp. 3d 386, 399 (S.D.N.Y. 2016), involved a breach of contract action relating to residential mortgage-backed securities ("RMBS"). The contract required the defendant to provide prompt notice to the plaintiffs if it discovered that certain warranties in the contract had been breached, so the issue became whether the defendant had knowledge of certain breaches. See id. at 425. The plaintiffs argued that the defendants failed to provide notice of certain breaches despite having knowledge of them due to willful blindness. See id.
The court found that the defendant was not willfully blind despite shutting down its surveillance group, which had allowed it to discover any widespread breaches of warranties, because it had an unrelated justification for shutting down the surveillance group. See id. at 426-27 (explaining that the surveillance group laid off "hundreds of people, almost [the] entire business in the Fall of 2007, and then essentially everyone else was laid off in the Spring of 2008 .... the entire business was imploding"). The court reasoned that "the principal purpose of the surveillance group was to monitor loans with an eye toward the marketing of future RMBS products to investors.... The closing of the surveillance group ... dovetailed with the collapse of the broader housing market and investor interest in RMBS deals." Id. at 427. Thus, the court concluded that the closing of the surveillance group "does not prove that [defendant] took ‘deliberate actions to avoid confirming a high probability of wrongdoing,’ or that [defendant] ‘can almost be said to have actually known the critical facts’ concerning widespread breaches across the three trusts." Id. (quoting Scher L. Firm, 948 N.Y.S.2d at 338 ) (citing Viacom Int'l, 676 F.3d at 35 ).
The procedural posture of U.S. Bank is most analogous to the instant case given that the opinion was also written after conducting a bench trial. See 205 F. Supp. 3d 386 at 399.
Therefore, New York's willful blindness doctrine imposes upon a party a duty to inquire, the scope of which is defined by the extent of the knowledge of the breach that it possesses. In other words, the more a party knows about the existence and extent of the breach, the more direct and specific is the duty to inquire. See Reed v. Gannon, 50 N.Y. 345, 349-51 (1872) ("The omission to make the proper examination or inquiry is gross negligence .... If the natural and obvious inquiry was not even made ... it was one of the cases characterized in the authorities as willful blindness." (emphasis added)).
b. Applying the Doctrine to the Instant Case
With this background in mind, the issue here is whether Nuance obtained actual knowledge prior to June 30, 2014, that IBM was not providing it with DeepQA updates developed by IBM Software as required under the agreement. The statute of limitations is an affirmative defense. Therefore, for IBM to prevail, it must show by a preponderance of the evidence that before June 30, 2014, Nuance obtained such actual knowledge. In this case, actual knowledge can be inferred from the evidence that Nuance knew that (1) IBM Software was developing updates to DeepQA and (2) IBM was withholding code developed by IBM Software.
As to the first prong (i.e., that Nuance knew IBM Software was developing updates to DeepQA), the following evidence shows Nuance knew IBM Software was creating updates to DeepQA and products based on DeepQA:
• David McQueeney (IBM's Chief Innovation Officer and former Vice President of Software in IBM Research) credibly testified that he told Nuance that "IBM Research had given [IBM Software] a copy of DeepQA, and [IBM Software] was working on commercializing this technology." McQueeney Decl. 5.
• In a 2012 email between several Nuance and IBM employees, Angus McIntyre (a Group Product Manager for IBM) notes "our [IBM Software] efforts on Watson: 1) Our specific enhancements (completed and planned) to the DeepQA engine and 2) Our specific products in the analytic space." DX71.
• In a 2012 internal Nuance email, Joseph Petro (Nuance's Chief Technology Officer) said Angus McIntyre told him that "[t]he IBM [Software] is working specifically on the Jeopardy engine and generalizing to a base core set of capabilities that are domain agnostic." DX88.
• Joseph Petro also credibly testified that "by the latest, April 3rd, 2012,
[he was] aware that the IBM software group was working specifically on DeepQA and, in fact, the domain independent code associated with DeepQA." Trial Tr. 92–93.
• In a 2012 email from Jeanne McCann (Nuance's former Chief Information Officer) to Paul Ricci, Nuance's CEO at the time, McCann noted "they have done work on ‘vanilla’ productization of Watson, solely in the IBM [Software]. They took an image of the code and have gone on their own." DX87.
• McCann also credibly testified by way of deposition that Nuance was aware of "software hygiene or productization" work "that the software group had done," which she defined as being "the kinds of work you might do to make sure that you can apply it to a number of third-party kinds of customers." McCann Dep. 188.
• In a 2011 email to Ricci, McCann noted "IBM [Software's] investments in ‘industrialization’ of the Watson domain-independent core." DX138.
This testimony, as well as all other deposition testimony referred to in this memorandum, was made part of the trial record.
The Court finds by a preponderance of the evidence that Nuance, at the highest level of management, knew IBM was creating updates to DeepQA.
As to the second prong (i.e., Nuance's knowledge that IBM was withholding code developed by IBM Software), the following evidence shows Nuance knew IBM would not provide it with IBM Software's DeepQA work:
• McQueeney (IBM's Chief Innovation Officer and former Vice President of Software in IBM Research) credibly testified that Ricci was told at a meeting "that Nuance was not entitled to any Modifications developed by [IBM Software]." McQueeney Decl. 7.
• McCann credibly testified by way of deposition that she understood Manoj Saxena, IBM's General Manager for the Watson Division, to be threatening "that IBM wasn't going to deliver the full set of updates to the DeepQA domain-independent core" prepared "[o]utside of research." McCann Dep. 166–67.
• A 2013 internal memorandum by McCann to Ricci noted the IBM Research-specific nature of the updates in saying that McQueeney "reinforced last week that we were indeed receiving all the work IBM Research was doing on Watson." DX103 (emphasis added).
• John Kelly (IBM's Executive Vice President and former Senior Vice President of IBM Research) testified that starting in 2011, he repeatedly told Ricci that "IBM was not giving Nuance the work done by [IBM Software] to commercialize DeepQA." Kelly Decl. 10.
Nuance impeached Kelly during cross examination with (1) his admission that "there are no documents in this case that back[ ] up what [he] claim[s] [he] told [Ricci]," July 28 Trial Tr. 56–57, and (2) contemporaneous internal Nuance emails from 2010 where Ricci says the "head of [IBM] research brought up ... the application of this technology to the medical field," PX7, and "made very strong commitments concerning resource dedication and rapid focus on achieving a demonstrable product in healthcare," PX24. See July 28 Trial Tr. 37–38 (Kelly testified that his "recollection of the conversation back in September of 2010 is different than ... what is marked in this contemporaneous document."). That said, Michael Rhodin, IBM's former Senior Vice President of Watson Business Development, testified in his deposition consistently with Kelly—i.e., that he also told Ricci that IBM Software was "going to create a commercialization project" and that "Nuance would not get the updates or upgrades made to the software." Rhodin Dep. 73–74. The Court finds this testimony credible.
• Credible deposition testimony from Mark Fanty (Nuance's Director of Core Research) indicates that Murthy Devarakonda (a Distinguished Research Staff Member and Manager for IBM) told him that Nuance would get "only what's from research not from what's outside research," and that "[y]ou don't get anything from software group," despite Fanty's knowledge "that the software group was doing something that involved DeepQA." Fanty Dep. 43, 83, 108.
• A 2013 internal memorandum written by McCann to Ricci states as follows:
the IBM Software Group wants to be absolutely sure that none of their software or work is to be made available to Nuance. There may well have been a transfer of the entire Watson code tree to the Software Group in order to allow IBM to execute work in the [Software Group] completely independently of IBM Research in order to keep any such work exclusive to IBM.
DX 103.4.
• McCann noted in a 2011 internal email that "we need to have our attorneys involved at this point on the whole updates thing." DX84.
• Another internal Nuance email from McCann to Ricci notes "a notable ‘pulling back’ by IBM" on, inter alia, giving Nuance "[a]ccess to the ‘full’ set of modifications by IBM to Watson Core." DX138.6.
• Yet another internal Nuance email from McCann to Ricci stated, "And Watson of course - where IBM Research may be adhering to the letter of any agreements but failed the spirit." DX126.
Moreover, on this second prong, IBM told Nuance that a firewall had been erected between IBM Research and IBM Software for the purpose of preventing Nuance from accessing IBM Software work product:
• Kelly testified he told Ricci that "IBM had erected a virtual firewall [to prevent] Nuance claiming [IBM Software]’s DeepQA commercialization work." Kelly Decl. 10.
• McCann's credible deposition testimony said that Nuance knew "that IBM Software group was not having IBM Research collaborate directly on its projects," and that this resulted in a suspicion "that Software group [was] doing work on the Watson code that [Nuance was] not getting." McCann Dep. 199–200, 214.
• A 2013 internal memorandum written by McCann to Ricci states, "Due to their interpretation [of the agreement] (or concern that there may be a different interpretation), the IBM Software Group is not allowing any joint work with IBM Research to occur." DX103.4.
Therefore, the Court finds by a preponderance of the evidence that Nuance knew IBM was withholding code developed by IBM Software.
Given what IBM communicated to Nuance, and what Nuance understood at the time, the Court concludes that Nuance was sufficiently aware of the significant possibility that updates to DeepQA were being developed by IBM Software, and that not all the updates would be delivered to Nuance, specifically including the key software hygiene and productization updates in the "blue-washed" code. See U.S. Bank, 205 F. Supp. 3d at 425 ("[A] willfully blind defendant is one who ... can almost be said to have actually known the critical facts." (quoting Scher L. Firm, 948 N.Y.S.2d at 338 ) (citing Viacom Int'l, 676 F.3d at 35 )). Despite this awareness, Nuance avoided confirming the information it received from IBM by making only general and non-specific inquiries concerning the updates that were not being delivered to it. See id. at 427 (describing willful blindness as taking "deliberate actions to avoid confirming a high probability of wrongdoing" (quoting Scher L. Firm, 948 N.Y.S.2d at 338 ) (citing Viacom Int'l, 676 F.3d at 35 )).
Thus, Nuance's argument that it could not independently verify what updates were being provided and what updates were not being provided is not persuasive, given that Nuance could have confirmed the breach by asking IBM more specific questions.
Notwithstanding the above, Nuance claims that upon receiving some information that IBM was not performing under the agreement, it sought general assurances from IBM that it was providing all the updates to which Nuance was entitled under the agreement. Nuance argues it did not seek specific assurances because it thought it might get the updates anyway or that the withholding of IBM Software updates was only temporary. The key question is then whether Nuance's general and non-specific inquiries to IBM were sufficient to satisfy its duty to inquire, arising from the information it had in its possession at the time—i.e., that IBM was not providing it with all the DeepQA updates under the agreement. Or, to put it another way, as IBM contends, did Nuance consciously avoid confirming that IBM was in breach of the agreement by making only general and non-specific inquiries of IBM? See id. at 425 ("Under New York law, where a party ‘had sufficient information to impose a duty upon it to make further inquiry ... its failure to do so constituted "willful blindness." ’ " (quoting Scher L. Firm, 948 N.Y.S.2d at 337 )).
The following evidence shows that Nuance consciously avoided confirming that IBM was not providing it with DeepQA updates developed by IBM Software by not asking IBM that specific question but only seeking general assurances, despite the fact that Nuance could not independently verify what updates were being provided and what updates were not being provided:
• In a 2013 email correspondence between Nuance and IBM, Nuance's Mark Fanty stated that one of the things Nuance wanted "to get straight going forward" was "the scope of the domain independent work that Nuance will continue to receive." PX84.2. IBM's Murthy Devarakonda responded and stated that moving forward "under the [agreement], we will deliver Modifications to the Watson code as per the [agreement]," and proceeded to quote the entire updates provision to the agreement. PX84.
• Ricci testified that given "a growing concern within Nuance ... that they weren't really sharing everything with us ... I would bring this up with [Kelly]." Trial Tr. 522.
• As McCann acknowledged, "Nuance [n]ever reach[ed] out to IBM to say, you know, are you providing us with all of the productization updates that the software group is making to the domain-independent core," but instead Nuance "assumed [IBM would] be living up to the agreement." McCann Dep. 196-97.
• McCann told McQueeney that Nuance "want[s] to be assured that we are receiving all of the updates to which we're entitled." McCann Dep. 239.
• McQueeney confirmed that McCann "question[ed] [him] to make sure [Nuance] received all of the updates [Nuance] was entitled to." Trial Tr. 299-300.
• Paul Reardon (former IBM Business Development Manager for IBM Research) similarly testified that he "heard some complaints from Nuance that they believed they were not receiving sufficient updates pursuant to the [agreement]." Trial Tr. 397.
• Ricci testified that "Kelly assured [him] that IBM was working on progressing and developing DeepQA, IBM was still very committed to DeepQA, but things take time, and IBM was doing everything they said they would do, including providing Nuance all IBM updates, upgrades and other changes and modifications to the DeepQA technology." Ricci Decl. 10.
• Helgi Bloom (Nuance's former Vice President of Corporate Development) testified that Nuance, mostly through McCann, "sought reassurance from IBM and received it." Trial Tr. 239-40.
• McQueeney testified that he assured "McCann that Nuance had received everything it was entitled to." Trial Tr. 300.
• Deposition testimony from McCann corroborates that McQueeney gave her "assurances that the code base was the same [as IBM Research was using] and that we were receiving all the updates." McCann Dep. 184, 235.
Given that Nuance knew (1) IBM Software was developing products based on DeepQA and (2) IBM was withholding code developed by IBM Software, the aforementioned generalized inquiries by Nuance and assurances by IBM were insufficient to satisfy Nuance's duty to inquire. The scope of the duty to inquire, as explained above, is defined by the extent of the knowledge that the non-breaching party has in its possession. See supra Section IV.B.1.a.
Under these circumstances, the Court concludes that Nuance's awareness of IBM Software's work on DeepQA and IBM's withholding of IBM Software's work from Nuance means it "had sufficient information to impose a duty upon it to make further inquiry" under the willful blindness doctrine. See U.S. Bank, 205 F. Supp. 3d at 425 (quoting Scher L. Firm, 948 N.Y.S.2d at 337 ). And by only asking in general terms and not directly or specifically about the IBM Software updates being withheld, the "failure to do so constituted ‘willful blindness.’ " Id. (quoting Scher L. Firm, 948 N.Y.S.2d at 337 ). Unlike in U.S. Bank, where there was a justification unrelated to willful blindness for why the defendant was unable to learn of the breach, see supra Section IV.B.1.a, there is no plausible explanation here for Nuance's failure to make direct and specific inquiries to IBM. Even if, as Nuance contends, it believed it would get the IBM Software updates later despite their being withheld at the time, this does not excuse its failure to inquire in these circumstances.
Given Nuance's willful blindness, the Court further concludes that Nuance had actual knowledge that IBM was breaching the agreement prior to June 30, 2014, i.e., more than two years before Nuance filed suit. Thus, Nuance's claims are barred by the contractual limitations period because it had actual knowledge of the breach more than two years before it filed suit.
***
Nuance's arguments that the continuing breach doctrine or the doctrine of equitable estoppel saves their claims are unavailing and will be explained seriatim.
2. Continuing Breach
The continuing breach doctrine does not protect Nuance's claims from being barred by the contractual limitations period.
Under the continuing breach doctrine, where "a contract requires continuing performance over a period of time, each successive breach may begin the statute of limitations running anew." Guilbert v. Gardner, 480 F.3d 140, 150 (2d Cir. 2007) (first citing Bulova Watch Co. v. Celotex Corp., 46 N.Y.2d 606, 415 N.Y.S.2d 817, 389 N.E.2d 130, 132 (1979) ; then citing Stalis v. Sugar Creek Stores, Inc., 295 A.D.2d 939, 744 N.Y.S.2d 586, 587-88 (2002) ; then citing Orville v. Newski, Inc., 155 A.D.2d 799, 547 N.Y.S.2d 913, 914 (1989) ; and then citing Airco Alloys Div. v. Niagara Mohawk Power Corp., 76 A.D.2d 68, 430 N.Y.S.2d 179, 186 (1980) ). Thus, each successive breach "re-commences the statute of limitations such that damages can be awarded beginning ‘from the date calculated by subtracting the limitations period from the date of filing.’ " Kwan v. Schlein, 441 F. Supp. 2d 491, 501 (S.D.N.Y. 2006) (quoting West Haven v. Com. Union Ins. Co., 894 F.2d 540, 546 (2d Cir. 1990) ; and relying on continuing breach principles under New York law, see Won's Cards, Inc. v. Samsondale/Haverstraw Equities, Ltd., 165 A.D.2d 157, 566 N.Y.S.2d 412, 415 (1991) ).
However, the doctrine only applies where there is "a series of independent, distinct [breaches]," not one breach that continues to cause harm or increases the damages by the passage of time without performance. See Mindspirit, LLC v. Evalueserve Ltd., 346 F. Supp. 3d 552, 593 (S.D.N.Y. 2018) (quoting Henry v. Bank of Am., 147 A.D.3d 599, 48 N.Y.S.3d 67, 70 (2017) ). In other words, it is insufficient that a single breach causes multiple effects; instead, "the plaintiff's claims must be ‘inherently susceptible to being broken down into a series of independent and distinct events or wrongs, each having its own associated damages.’ " Miele v. Pension Plan of N.Y. State Teamsters Conf. Pension & Ret. Fund, 72 F. Supp. 2d 88, 102 (quoting Brown Park Estates-Fairfield Dev. Co. v. United States, 127 F.3d 1449, 1456 (Fed. Cir. 1997) ); see also Selkirk v. State, 249 A.D.2d 818, 671 N.Y.S.2d 824, 825 (1998) ("[A]pplication of the doctrine ... may only be predicated on continuing unlawful acts and not on the continuing effects of earlier unlawful conduct." (citation omitted)).
A contract that requires continuous performance may be breached partially a series of times or completely a single time, and the latter, a single complete breach, is effected by repudiation or by a failure to perform that substantially frustrates the purpose of the contract. Kwan, 441 F. Supp. 2d at 501 (quoting 4 Corbin on Contracts, ch. 53, § 956 (1951)) (citing Won's Cards, 566 N.Y.S.2d at 415 ); see Kobayashi Ventures, LLC v. Papertech Inc., No. 08 Civ. 04450, 2009 WL 10738379, at *4 (S.D.N.Y. Feb. 23, 2009) ("Where the continuous performance required by a contract contemplates something more than cash payments, the statute of limitations begins to run when nonperformance is coupled with repudiation." (citing Restatement (Second) of Contracts § 243(2) (Am. L. Inst. 1981), and relying on principles of continuous performance under New York law, see, e.g., Green v. Petersen, 218 N.Y. 280, 112 N.E. 746 (1916) )). Repudiation occurs "when a party manifests an intent not to perform, either by words or by deeds." Aetna Cas. & Sur. Co. v. Aniero Concrete Co., 404 F.3d 566, 587 (2d Cir. 2005) (first, third, and fourth citations omitted) (citing Key Bank of N.Y., N.A. v. K.H. Assocs., 210 A.D.2d 769, 620 N.Y.S.2d 537, 537 (1994) ).
Here, the continuing breach doctrine applies because the contract requires continuing performance over ten years. However, the doctrine does not preserve Nuance's claims because IBM's breach was a single complete breach. Nuance claims that IBM breached the contract by not giving Nuance all the updates connected to the DeepQA technology. But as explained further in the Court's breach of contract discussion, see supra Section IV.A, the Court finds that IBM breached a material term of the agreement once—when it failed to give Nuance the set of updates that were "blue-washed" (i.e., the process of deleting code that is not suitable for commercial use and replacing it with code that can be used in a commercial product).
Regardless of whether IBM delivered or failed to deliver any of the other updates, the withholding of the blue-washed code materially breached the agreement and substantially frustrated its purpose. See Kwan, 441 F. Supp. 2d at 501 (quoting 4 Corbin on Contracts, ch. 53, § 956 (1951)) (citing Won's Cards, 566 N.Y.S.2d at 415 ) (explaining that a single complete breach is effected by either repudiation or by a failure to perform that substantially frustrates the purpose of the contract). Thus, because IBM's breach was a single complete breach, the doctrine of continuing breach does not defeat the statute of limitations defense.
Here, there was no repudiation because IBM did not specifically and unequivocally tell Nuance that Nuance would not receive any updates developed outside of IBM Research. Nonetheless, there was a single complete breach because IBM's failure to provide Nuance with the IBM Software updates that made DeepQA commercially usable, in 2012, was a failure to perform that substantially frustrated the purpose of the contract.
Alternatively, Nuance did not show by the preponderance of the evidence that after June 30, 2014, IBM developed non-product updates that made DeepQA commercially viable. See supra Sections III.D.2, III.E, IV.A.2. Its continuing breach argument focuses on IBM's post-2014 development of products—e.g., Watson for Oncology—the withholding of which by IBM was not a breach under the SLA. See supra Section IV.A.2. Therefore, the breach Nuance proved occurred before the applicable statute of limitations period.
3. Equitable Estoppel
Equitable estoppel likewise does not preserve Nuance's claims because Nuance's knowledge of the scope and extent of the breach precludes the application of the doctrine. Under the doctrine of equitable estoppel, a court may "bar the application of the statute of limitations due to representations or conduct by a party which induces the opposing party to postpone bringing suit on a known cause of action, or which fraudulently conceals an action unknown to the opposing party." Abercrombie v. Andrew Coll., 438 F. Supp. 2d 243, 265 (S.D.N.Y. 2006) (citing Knaysi v. A.H. Robins Co., 679 F.2d 1366, 1368 (11th Cir. 1982) (quoting Parsons v. Dep't of Transp., 74 Misc.2d 828, 344 N.Y.S.2d 19, 24 (Sup. Ct. 1973) )).
But equitable estoppel does not apply "where a plaintiff possesses ‘timely knowledge sufficient to place him or her under a duty to make inquiry and ascertain all the relevant facts prior to the expiration of the applicable Statute of Limitations.’ " Id. at 266 (quoting Gleason v. Spota, 194 A.D.2d 764, 599 N.Y.S.2d 297, 298 (1993) ). Additionally, "equitable estoppel will preclude a defendant from using the statute of limitations as a defense ‘where it is the defendant's affirmative wrongdoing ... which produced the long delay between the accrual of the cause of action and the institution of the legal proceeding.’ " Putter v. N. Shore Univ. Hosp., 7 N.Y.3d 548, 825 N.Y.S.2d 435, 858 N.E.2d 1140, 1142 (2006) (citation omitted). A plaintiff relying on equitable estoppel holds the burden to establish their timeliness was prevented by a defendant's specific actions of fraud, deception, or misrepresentation. See, e.g., Simcuski v. Saeli, 44 N.Y.2d 442, 406 N.Y.S.2d 259, 377 N.E.2d 713, 717 (1978).
The equitable estoppel doctrine does not apply here because (1) Nuance obtained actual knowledge prior to June 30, 2014, and (2) while IBM provided Nuance with several assurances of performance, such representations do not rise to the level of fraud, deception, or misrepresentation.
IBM's representations do not rise to the level of fraud, deception, or misrepresentation because Nuance has not shown that IBM gave its assurances with intent to defraud, knowledge of their falsity, or disregard for their truth. See Restatement (Second) of Contracts § 162 (Am. L. Inst. 1981) ; 26 Willison on Contracts § 69:3 (4th ed. 2021); see also Simcuski, 406 N.Y.S.2d 259, 377 N.E.2d at 719 ("With respect to the application of the doctrine of equitable estoppel to a defense of Statute of Limitations ... we are concerned with an intentional, not merely negligent, wrong ....").
As a result of the foregoing, Nuance's claims related to updates developed by IBM Software are barred by the contractual limitations period.
V. SUMMARY
For the reasons set forth above, the Court finds that Nuance was entitled to receive one set of updates that would allow the DeepQA technology to be used in a commercial setting (i.e., the blue-washed code), and while IBM delivered some updates to Nuance, IBM breached the agreement by failing to deliver the blue-washed code. However, Nuance was sufficiently aware of the significant possibility that IBM was breaching the agreement prior to June 30, 2014 (i.e., more than two years before Nuance filed suit). Given what Nuance knew at the time, Nuance avoided making the appropriate inquiry to IBM and was willfully blind. Given Nuance's willful blindness, the Court further concludes that Nuance had actual knowledge that IBM was breaching the agreement prior to June 30, 2014. Under these circumstances, Nuance's claims are barred by the contractual limitations period. Accordingly, judgment shall be entered by separate document in favor of IBM and against Nuance.
JUDGMENT
IT IS HEREBY ORDERED that judgment is entered in favor of International Business Machines Corp. and against Nuance Communications, Inc. for the reasons explained in the accompanying memorandum.