`
`UNITED STATES DISTRICT COURT
`WESTERN DISTRICT OF WASHINGTON
`AT SEATTLE
`
`
`STEVEN VANCE, et al.,
`
`CASE NO. C20-1082JLR
`
`v.
`
`Plaintiffs,
`
`
`
`
`MICROSOFT CORPORATION,
`
`ORDER ON MICROSOFT’S
`MOTION FOR SUMMARY
`JUDGMENT
`
`
`
`
`Defendant.
`
`I.
`
`INTRODUCTION
`
`Before the court is Defendant Microsoft Corporation’s (“Microsoft”) renewed
`
`
`
`
`
`
`
`motion for summary judgment. (Mot. (Dkt. # 127); Reply (Dkt. # 138).) Plaintiffs
`
`Steven Vance and Tim Janecyk (collectively, “Plaintiffs”) oppose Microsoft’s motion.
`
`(Resp. (Dkt. # 1351).) The court has considered the motion, all materials submitted in
`
`
`1 Plaintiffs originally filed their response under seal because it relied on and cited
`documents that Microsoft had marked confidential; they also filed a redacted version of their
`response. (Mot. to Seal (Dkt. # 134); Redacted Resp. (Dkt. # 132).) Because Microsoft did not
`oppose unsealing the response and the documents, the court denied Plaintiffs’ motion to seal and
`
`ORDER - 1
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 2 of 22
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`support of and in opposition to the motion, and the governing law. Being fully advised,2
`
`the court GRANTS Microsoft’s motion for summary judgment.
`
`II.
`
`BACKGROUND
`
`
`
`The court sets forth the factual and procedural background of this case below.
`
`A.
`
`Factual Background
`
`
`
`
`
`1.
`
`The Diversity in Faces (“DIF”) Dataset
`
`Plaintiffs are longtime Illinois residents who, beginning in 2008, uploaded digital
`
`photographs, including photos of themselves, to Flickr, a photo-sharing website. (See
`
`Compl. (Dkt. # 1) ¶¶ 6-7, 28, 66-67, 75; Vance Dep.3 at 9:15-10:9; Janecyk Dep.4 at
`
`10
`
`39:7-40:1.) In 2014, Yahoo!, Flickr’s then-parent company, publicly released a dataset of
`
`11
`
`about 100 million photographs that had been uploaded to Flickr’s website between 2004
`
`
`directed the clerk to remove the seal on Plaintiffs’ responsive brief and the confidential
`documents. (Mot. to Seal Resp. (Dkt. # 136); 7/11/22 Order (Dkt. # 137).) Accordingly, the
`court cites the unredacted version of Plaintiffs’ response in this order.
`
` 2
`
` Both parties request oral argument on the motion (see Mot. at 1; Resp. at 1). The court,
`however, concludes that oral argument would not be helpful to its disposition of the motion. See
`Local Rules W.D. Wash. LCR 7(b)(4).
`
` 3
`
` Both parties have submitted excerpts from Mr. Vance’s deposition. (See Berger Decl.
`(Dkt. # 86) ¶ 2, Ex. 1; 7/1/22 Lange Decl. (Dkt. # 132-1) ¶ 2, Ex. 1.) For ease of reference, the
`court cites directly to the page and line number of the deposition.
`The court notes that Plaintiffs did not highlight the portions of the deposition transcripts
`that they referred to in their pleadings as required by Local Civil Rule 10(e)(10). See Local
`Rules W.D. Wash. LCR 10(e)(10) (“All exhibits [submitted in support of or in opposition to a
`motion] must be marked to designate testimony or evidence referred to in the parties’ filings.”).
`The court directs Plaintiffs’ counsel to review the local rules regarding marking exhibits before
`making any further filings.
`
` 4
`
` Both parties have submitted excerpts from Mr. Janecyk’s deposition. (See Berger Decl.
`¶ 3, Ex. 2; 7/1/22 Lange Decl. ¶ 3, Ex. 2.) For ease of reference, the court cites directly to the
`page and line number of the deposition.
`
`ORDER - 2
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 3 of 22
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`and 2014 (the “YFCC-100M Dataset”). (See Merler Decl. (Dkt. # 85) ¶ 3, Ex. A
`
`(“Diversity in Faces”) at 2.) The YFCC-100M Dataset included photos uploaded by both
`
`Plaintiffs. (See Vance Dep. at 179:22-23; Janecyk Dep. at 95:22-24.)
`
`
`
`Before 2018, “there was an industry-wide problem with many facial recognition
`
`systems’ ability to accurately characterize individuals who were not male and did not
`
`have light colored skin tones.” (Merler Decl. ¶ 4.) As a result, “the facial recognition
`
`systems and algorithms associated with those facial recognition systems were trained in
`
`such a way that the systems were able to accurately characterize a white, light skinned
`
`male subject, but the technology suffered from inaccuracies when it had to characterize a
`
`non-male or a person with darker skin tones.” (Id.) Seeking to “advance the study of
`
`fairness and accuracy in face recognition technology,” researchers working for
`
`12
`
`International Business Machines Corporation (“IBM”)5 used one million of the photos in
`
`13
`
`the YFCC-100M Dataset to develop the Diversity in Faces (“DiF”) Dataset at issue in
`
`14
`
`this case. (Id. ¶ 5; Diversity in Faces at 2, 7.) The researchers implemented ten “facial
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`
`
`coding schemes” to measure aspects of the facial features of the individuals pictured in
`
`the photos, such as “craniofacial distances, areas and ratios, facial symmetry and contrast,
`
`skin color, age and gender predictions, subjective annotations, and pose and resolution.”
`
`(Diversity in Faces at 9.) A statistical analysis of these coding schemes “provided insight
`
`into how various dimensions . . . provide indications of dataset diversity.” (Merler
`
`
`5 All of the researchers involved in creating the DiF Dataset were based in and worked
`out of IBM’s office in Yorktown Heights, New York; and the work was performed on and stored
`on IBM Research computer servers in Poughkeepsie, New York. (Id. ¶ 8.) None of the work
`involved computers or systems located in Illinois. (Id.)
`
`ORDER - 3
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 4 of 22
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`Decl. ¶ 6.) The coding schemes implemented by the IBM researchers were intended to
`
`enable other researchers to develop techniques to estimate diversity in their own datasets,
`
`with the goal of mitigating dataset bias, and were “never intended to identify any
`
`particular individual.” (Id. ¶ 7.) Rather, the coding schemes were “purely descriptive
`
`and designed to provide a mechanism to evaluate diversity in the dataset.” (Id.)
`
`IBM provided the DiF Dataset free of charge to researchers who filled out a
`
`questionnaire and submitted it to IBM via email. (Id. ¶¶ 4, 9.) The questionnaire
`
`required the researcher to verify
`
`(i) that he/she would only use the DiF Dataset for research purposes, and
`(ii) that he/she had read and agreed to the DiF Dataset terms of use, which
`made clear that the DiF Dataset could only be used for non-commercial,
`research purposes and prohibited using the DiF Dataset to identify any
`individuals in images associated with URLs in the DiF Dataset.
`
`(Id. ¶ 9; see also id. ¶ 11, Ex. H (DiF Dataset terms of use).) After verifying that a
`
`request was for a “legitimate research purpose,” IBM researcher Dr. Michele Merler sent
`
`the DiF Dataset to the requesting researcher “via an email that included a link to a
`
`temporary Box folder that contained the DiF Dataset.” (Merler Decl. ¶ 10.)
`
`2.
`
`Plaintiffs’ Photos in the DiF Dataset
`
`17
`
`
`
`The DiF Dataset includes at least 61 of the nearly 19,000 public photos that Mr.
`
`18
`
`Vance uploaded to Flickr. (Vance Dep. at 179:22-23, 210:19-24.) Mr. Vance appears in
`
`19
`
`some of the photos in the DiF Dataset; other photos depict people whose state of
`
`20
`
`21
`
`22
`
`
`
`residence was unknown to Mr. Vance and at least one depicts individuals who themselves
`
`were unknown to Mr. Vance. (Id. at 132:4-14; 154:5-16.)
`
`ORDER - 4
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 5 of 22
`
`
`
`
`
`The DiF Dataset includes 24 of the 1,669 public photos that Mr. Janecyk uploaded
`
`to Flickr. (Janecyk Dep. at 74:21-24, 95:22-96:1.) Mr. Janecyk appears in at least one of
`
`the photos. (Id. at 99:21-100:6.) Because Mr. Janecyk photographed people on the
`
`streets of Chicago, however, he does not know the names or places of residence of the
`
`individuals depicted in most of his photos. (Id. at 45:16-46:19, 98:8-100:13,
`
`167:11-168:15, 228:19-21.)
`
`
`
`
`
`3. Microsoft’s Downloads of the DiF Dataset
`
`Two individuals affiliated with Microsoft downloaded the DiF Dataset in February
`
`2019: contractor Benjamin Skrainka and Microsoft Research intern Samira Samadi.
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`(Skrainka Decl. (Dkt. # 87) ¶ 5; Samadi Decl. (Dkt. # 88) ¶¶ 5-6.) The court describes
`
`11
`
`their interactions with the DiF Dataset below.
`
`12
`
`13
`
`14
`
`15
`
`
`
`
`
`
`
`a.
`
`Benjamin Skrainka
`
`Between September 7, 2018, and August 1, 2019, Mr. Skrainka worked as an
`
`independent contractor for Neal Analytics, LLC, a Washington-based consulting firm,
`
`through which he contracted as a vendor to Microsoft. (Skrainka Decl. ¶ 2; Skrainka
`
`16
`
`Dep.6 at 91:7-24, 111:8-23.) During this period, Mr. Skrainka provided support for a
`
`17
`
`project, Azure Intelligent Storage (“AIS”), for Microsoft. (Skrainka Decl. ¶ 3.) His work
`
`18
`
`related to defining a benchmark protocol for evaluating a third-party facial recognition
`
`19
`
`20
`
`21
`
`22
`
`
`
`technology that Microsoft was considering acquiring. (Id.; Kasap Decl. (Dkt. # 91)
`
`
`6 Both parties have submitted excerpts from Benjamin Skrainka’s deposition. (See
`5/19/22 Wiese Decl. (Dkt. # 129) ¶ 2, Ex. 1; 7/1/22 Lange Decl. ¶ 12, Ex. 11; 7/29/22 Wiese
`Decl. (Dkt. # 139) ¶ 2, Ex. 9.) For ease of reference, the court cites directly to the page and line
`number of the deposition.
`
`ORDER - 5
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 6 of 22
`
`
`
`¶¶ 2-3.) Mustafa Kasap, a Principal Program Manager, was Mr. Skrainka’s manager and
`
`technical advisor for this project. (Skrainka Decl. ¶ 4; Kasap Decl. ¶ 2; Skrainka Dep. at
`
`121:11-15.) As part of his project, Mr. Skrainka “determined what the parameters and/or
`
`methodology should be for comparing different face recognition technologies available in
`
`the market,” including the technology that Microsoft considered acquiring, and developed
`
`code to implement these benchmarks. (Skrainka Decl. ¶ 3; Kasap Decl. ¶ 4.) He
`
`researched datasets containing photographs that might be suitable for making these
`
`comparisons. (Skrainka Decl. ¶ 4.)
`
`On or about February 1, 2019, Mr. Skrainka requested a copy of the DiF Dataset
`
`from IBM. (Skrainka Decl. ¶ 4.) After Mr. Skrainka filled out IBM’s questionnaire,
`
`IBM granted him access to the DiF Dataset through an online link. (Skrainka Decl. ¶ 5.)
`
`He downloaded the DiF Dataset sometime in early February 2019 while in Washington.
`
`(Id.) Mr. Skrainka evaluated the suitability of the photographs linked in the DiF Dataset
`
`for his project by manually inspecting some of the images and used some of the metadata
`
`in the DiF Dataset to “shrink or expand images to a consistent size” and to “extract the
`
`relevant faces” before running Microsoft’s facial recognition software on them.
`
`(Id. ¶¶ 6-7; Skrainka Dep. at 229:18-230:21, 231:13-18, 233:15-235:6.) He was not
`
`interested in the coding schemes or facial annotations included in the DiF Dataset.
`
`(Skrainka Dep. at 227:21-229:23.) He determined that the images were not suitable for
`
`his benchmarking research because they “did not look like a conventional head-on
`
`photograph used on a driver’s license or passport” and were of generally low quality.
`
`(Skrainka Decl. ¶ 7; see also Skrainka Dep. at 233:7-235:6 (explaining reasons the
`
`ORDER - 6
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 7 of 22
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`images were not suitable).) Consequently, he did not further pursue using the DiF
`
`Dataset. (Skrainka Decl. ¶ 7.)
`
`Mr. Skrainka was unaware that the DiF Dataset included any photographs or data
`
`related to Illinois residents. (Id. ¶ 12.) He did not share the DiF Dataset with anyone at
`
`Microsoft and is unaware of any other Microsoft group using the DiF Dataset, although
`
`he acknowledges that others could have accessed the dataset while it was stored in the
`
`cloud without his knowing. (Id.; Skrainka Dep. at 362:14-363:24; see also Kasap Decl.
`
`¶ 7 (stating he, too, was unaware of any other Microsoft group using the DiF Dataset).)
`
`Mr. Skrainka does not recall where, “if at all,” he saved his copy of the DiF
`
`10
`
`Dataset. (Skrainka Decl. ¶ 8.) During his project, however, Microsoft instructed him to
`
`11
`
`use a virtual machine7 for his work, and he recalled that “any facial-recognition-related
`
`12
`
`work that [he] performed . . . was loaded only onto virtual machines and cloud storage in
`
`13
`
`Azure.” (Id.; Skrainka Dep. at 188:12-23.) When an Azure user creates a virtual
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`
`
`machine or sets up “blob” storage, he or she is prompted to select an Azure Region that
`
`determines the geography of the data centers where the data will be stored; each Azure
`
`Region includes “availability zones” that map to specific datacenters within the selected
`
`region. (Kuttiyan Decl. (Dkt. # 128) ¶ 3.) Mr. Skrainka believes that he used a “West
`
`Coast availability zone” for the work that he performed and that he was “almost surely”
`
`using West Coast data centers “because they’re faster.” (Skrainka Dep. at 147:2-20,
`
`
`7 “A virtual machine emulates the characteristics of a stand-alone physical computer. It
`shares physical resources, such as servers, with other virtual machines, and each virtual machine
`is isolated by software.” (Id.)
`
`ORDER - 7
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 8 of 22
`
`
`
`154:10-20.) He acknowledged, however, that he might have saved data in other
`
`availability zones, including the East Coast availability zone, and that he was unaware of
`
`whether the data was backed up or migrated to other availability zones. (Skrainka Dep.
`
`at 148:15-17, 151:19-24, 185:15-20; see also 7/1/22 Lange Decl. ¶ 19, Ex. 18 (“Kuttiyan
`
`Dep.”) at 62:17-63:4, 63:24-65:2 (acknowledging that data may be backed up to data
`
`centers in other availability zones).) When his project ended, Mr. Skrainka
`
`decommissioned all virtual machines that he used in the project, including any data stored
`
`on those virtual machines or in the cloud. (Skrainka Decl. ¶ 8.)
`
`Microsoft has been unable to confirm if and where on its systems Mr. Skrainka
`
`stored his copy of the DiF Dataset. (Mot. at 7; see Brunke Decl. (Dkt. # 92) ¶¶ 5-6.) In
`
`its July 15, 2021 supplemental answers to Plaintiffs’ interrogatories, it stated that if Mr.
`
`Skrainka’s copy of the DiF dataset had been stored on Microsoft servers in the cloud, the
`
`file would have been “chunked (i.e., divided into non-overlapping packets of data bits)”
`
`and encrypted, and the encrypted chunks would have been stored in data centers, likely in
`
`San Antonio, Texas and Chicago, Illinois. (Lange Decl. ¶ 20, Ex. 19 at 9-10
`
`(supplemental answer to ROG No. 8).) In its second supplemental answers served after
`
`Mr. Skrainka’s deposition, however, Microsoft “amend[ed] and correct[ed]” its
`
`supplemental answer to state that data stored on virtual machines and “blob” storage in
`
`the “West US” availability zone in February 2019 would have been stored in servers
`
`located in Washington or California, rather than in Illinois. (Id. at 10-11 (second
`
`supplemental answer to ROG No. 8); see Kuttiyan Decl. ¶ 4, Ex. A.)
`
`ORDER - 8
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 9 of 22
`
`
`
`
`
`
`
`
`
`b.
`
`Samira Samadi
`
`Between January 22, 2019, and May 3, 2019, while she was a graduate student at
`
`Georgia Institute of Technology in Atlanta, Georgia, Ms. Samadi completed a student
`
`internship at Microsoft Research in New York City, New York. (Samadi Decl. ¶ 2.) The
`
`“focus of [her] internship was a research project involving the study of how humans
`
`interact with, use, and make decisions with facial recognition systems.” (Id. ¶ 3.) She
`
`“wanted to design a controlled human-subject experiment where participants were shown
`
`examples of images of faces and asked to judge the similarities of the faces in the images
`
`given the similarity score generated by an automatic facial recognition system.” (Id.)
`
`Her goal was “to measure how the humans’ judgments of face similarities are affected by
`
`perceived race, skin tone, and gender of the faces they are shown.” (Id.) Her research
`
`was not focused on identifying people through facial recognition systems and did not
`
`involve measuring facial geometry or features. (Id. ¶ 4.) Ms. Samadi’s research was
`
`supervised by Microsoft Senior Principal Researcher Jenn Wortman Vaughan, who was
`
`also based in New York. (Vaughan Decl. (Dkt. # 89) ¶¶ 1, 3.)
`
`To run her experiment, Ms. Samadi needed “multiple photographs of the same
`
`individual, all of which needed to be directly facing the camera or slightly angled.”
`
`(Samadi Decl. ¶ 4.) On or about February 20, 2019, she asked IBM for access to the DiF
`
`Dataset using her Georgia Institute of Technology credentials. (Id. ¶ 5, Ex. A (email
`
`requesting access); see also Vaughan Decl. ¶ 7, Ex. B (email thread discussing Ms.
`
`Samadi’s request for the DiF Dataset).) IBM directed Ms. Samadi to fill out its
`
`questionnaire, and after she did so, granted her access to the DiF Dataset through an
`
`ORDER - 9
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 10 of 22
`
`
`
`online link. (Samadi Decl. ¶ 6.) She downloaded the dataset on or about February 25,
`
`2019, while working at Microsoft Research in New York. (Id.)
`
`
`
`After she downloaded the DiF Dataset, Ms. Samadi “briefly reviewed” some of
`
`the photographs linked in that dataset. (Id. ¶ 7; see also Samadi Dep.8 at 20:5-7 (stating
`
`she reviewed the DiF Dataset for about half an hour).) She determined that the photos
`
`were not suitable for her project because there were not multiple photos of the same
`
`individual facing the camera. (Samadi Decl. ¶ 7.) She did not further review the images
`
`in the DiF Dataset. (Id.; see also id., Ex. C (email to Dr. Vaughan, stating that after
`
`“looking closely” at the “IBM data,” she found that it did not have multiple images for
`
`one person and that the images have “many different backgrounds”).) She did not use the
`
`DiF Dataset in her project and the DiF Dataset did not play any role in the development
`
`of the paper that she wrote with Dr. Vaughan about the results of her research. (Id. ¶ 11,
`
`Ex. E (research paper); Vaughan Decl. ¶¶ 7-8.)
`
`Ms. Samadi was “not aware of or interested in any facial annotations” in the DiF
`
`Dataset, did not review any such data, and did not share the link to download the DiF
`
`Dataset with anyone else. (Samadi Decl. ¶ 8.) She did not know that the DiF Dataset
`
`included photographs or data relating to Illinois residents. (Id. ¶ 9.) Neither Ms. Samadi
`
`nor Dr. Vaughn are aware of any other projects at Microsoft Research that used the DiF
`
`
`8 Both parties have submitted excerpts from Samira Samadi’s deposition. (See 5/19/22
`Wiese Decl. (Dkt. # 129) ¶ 3, Ex. 2; 7/1/22 Lange Decl. ¶ 22, Ex. 21.) For ease of reference, the
`court cites directly to the page and line number of the deposition.
`
`ORDER - 10
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`11
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 11 of 22
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`Dataset, nor are they aware of the DiF Dataset being used by any other group at
`
`Microsoft. (Id. ¶ 12; Vaughan Decl. ¶ 11.)
`
`Ms. Samadi believes, but is not certain, that she downloaded the DiF Dataset to
`
`her Microsoft Research laptop. (Samadi Decl. ¶ 6; Samadi Dep. at 38:20-39:6.) Under
`
`Microsoft’s Data Retention and Disposal Standard, data saved on Ms. Samadi’s
`
`Microsoft Research laptop was deleted within 180 days of the end of her internship.
`
`(Swann Decl. (Dkt. # 90) ¶ 9.) It is possible, however, that Ms. Samadi may have saved
`
`the DiF Dataset to her OneDrive account9 or that her laptop was automatically uploading
`
`information to OneDrive. (See Chirico Decl. (Dkt. # 93) ¶ 2, Ex. A10 (“3/11/19 Samadi
`
`10
`
`Email”) (email from Ms. Samadi to Mr. Chirico, in which she states that she works with
`
`11
`
`and must download “big data sets” and that her OneDrive account was full); Samadi Dep.
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`
`
`at 65:6-20, 57:1-19; Lange Decl. ¶ 20, Ex. 19 at 9 (supplemental answer to ROG No. 8)
`
`(stating that Microsoft’s investigation “has not established how Ms. Samadi initially
`
`downloaded or stored the IBM DiF Dataset”).) If Ms. Samadi’s copy of the DiF dataset
`
`was saved to Ms. Samadi’s OneDrive account, the file would have been “chunked (i.e.,
`
`divided into non-overlapping packets of data bits) and encrypted, and the encrypted
`
`chunks would have been stored in [Microsoft’s] data centers, likely in San Antonio,
`
`Texas and Chicago, Illinois.” (Lange Decl. ¶ 20, Ex. 19 at 10-11 (second supplemental
`
`answer to ROG No. 8).)
`
`
`9 OneDrive is Microsoft’s cloud file hosting and storage service. (Swann Decl. ¶ 4.) It
`allows Microsoft personnel to store their files and data in the cloud. (Id.)
`
`10 Plaintiffs also provided this email thread as an exhibit. (See Lange Decl. ¶ 25, Ex. 24.)
`
`ORDER - 11
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 12 of 22
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`In March 2019, Ms. Samadi reached out to Jeff Chirico, who provided information
`
`technology support for Microsoft Research in New York, to ask where she should store
`
`data related to her research project. (3/11/19 Samadi Email; Samadi Dep. at 50:24-51:12;
`
`Chirico Decl. ¶¶ 1-2.) Mr. Chirico instructed her to save her data on a “hidden share” on
`
`a Microsoft Research server located in New York. (3/11/19 Samadi Email; Chirico Decl.
`
`¶¶ 2-3.) This server is not backed up to any other server, and access to the server is
`
`restricted to Microsoft Research. (Chirico Decl. ¶ 3.) Microsoft later found a copy of the
`
`DiF Dataset and other data from Ms. Samadi’s internship on the Microsoft Research
`
`server in New York. (Lange Decl. ¶ 20, Ex. 19 at 8 (answer to ROG No. 8).)
`
`10
`
`B.
`
`Relevant Procedural Background
`
`11
`
`
`
`Plaintiffs filed their proposed class complaint in this action on July 14, 2020.
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`(Compl.) They brought claims against Microsoft for violations of two provisions of
`
`Illinois’s Biometric Information Privacy Act, 740 ILCS § 14/1, et seq. (“BIPA”), unjust
`
`enrichment, and injunctive relief. (Id. ¶¶ 93-122.) With respect to the BIPA violations,
`
`Plaintiffs alleged that Microsoft (1) violated BIPA § 15(b) by collecting and obtaining
`
`their biometric data without providing required information or obtaining written releases,
`
`and (2) violated BIPA § 15(c) by unlawfully profiting from Plaintiffs’ biometric data.
`
`(Id. ¶¶ 93-106.)
`
`19
`
`
`
`On September 14, 2020, Microsoft moved to dismiss Plaintiffs’ claims. (MTD
`
`20
`
`21
`
`22
`
`
`
`(Dkt. # 25).) On March 15, 2021, the court granted in part and denied in part Microsoft’s
`
`motion to dismiss. (3/15/21 Order (Dkt. # 43).) The court (1) granted Microsoft’s
`
`motion to dismiss Plaintiffs’ injunctive relief claim on the ground that injunctive relief is
`
`ORDER - 12
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 13 of 22
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`not a standalone cause of action; (2) denied Microsoft’s motion to dismiss Plaintiffs’
`
`BIPA § 15(b) claim, concluding that Plaintiffs had sufficiently alleged the elements of the
`
`claim; and (3) deferred ruling on Microsoft’s motion to dismiss Plaintiffs’ BIPA § 15(c)
`
`and unjust enrichment claims pending the receipt of supplemental briefing. (See
`
`generally id.) On April 14, 2021, after reviewing the parties’ supplemental briefing and
`
`hearing oral argument, the court dismissed Plaintiffs’ BIPA § 15(c) claim with leave to
`
`amend and denied Microsoft’s motion to dismiss Plaintiffs’ unjust enrichment claim.
`
`(See 4/13/21 Min. Entry (Dkt. # 46); 4/14/21 Order (Dkt. # 47).) Despite being granted
`
`leave to do so, Plaintiffs did not amend their BIPA § 15(c) claim. (See generally Dkt.)
`
`10
`
`
`
`Microsoft filed its original motion for summary judgment on December 10, 2021.
`
`11
`
`12
`
`13
`
`14
`
`(1st MSJ (Dkt. # 84).) On February 8, 2022, the court granted in part Plaintiffs’ motion
`
`for additional discovery pursuant to Federal Rule of Civil Procedure 56(d) and struck
`
`Microsoft’s original motion for summary judgment without prejudice. (2/8/22 Order
`
`(Dkt. # 118); see also Pls. 56(d) Mot. (Dkt. # 107).)
`
`15
`
`
`
`On May 19, 2022, Microsoft filed the instant renewed motion for summary
`
`16
`
`17
`
`18
`
`19
`
`judgment. (See Mot.) Subsequently, the parties agreed to a stipulated briefing schedule
`
`to allow Plaintiffs to obtain additional discovery. (5/27/22 Stip. (Dkt. # 130).) Thus, this
`
`motion became ripe for decision on July 29, 2022. (Id.)
`
`III. ANALYSIS
`
`20
`
`
`
`Microsoft argues that summary judgment on Plaintiffs’ claims is warranted
`
`21
`
`22
`
`
`
`because (1) BIPA cannot apply extraterritorially to its conduct outside of Illinois as a
`
`matter of Illinois law; (2) applying BIPA to Microsoft’s conduct would violate the
`
`ORDER - 13
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 14 of 22
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`dormant Commerce Clause of the United States Constitution; (3) even if BIPA could
`
`apply to Microsoft’s out-of-state conduct, Plaintiffs cannot prove the elements of their
`
`BIPA § 15(b) claim; and (4) Plaintiffs cannot prove the elements of their unjust
`
`enrichment claim. (See generally Mot.) Below, the court sets forth the standard for
`
`evaluating motions for summary judgment before considering Microsoft’s motion.
`
`A.
`
`Summary Judgment Standard
`
`Under Rule 56 of the Federal Rules of Civil Procedure, either “party may move
`
`for summary judgment, identifying each claim or defense—or the part of each claim or
`
`defense—on which summary judgment is sought.” Fed. R. Civ. P. 56. Summary
`
`10
`
`judgment is appropriate if the evidence, when viewed in the light most favorable to the
`
`11
`
`non-moving party, demonstrates “that there is no genuine dispute as to any material fact
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`and the movant is entitled to judgment as a matter of law.” Id.; see Celotex Corp. v.
`
`Catrett, 477 U.S. 317, 322 (1986). A dispute is “genuine” if “the evidence is such that a
`
`reasonable jury could return a verdict for the nonmoving party.” Anderson v. Liberty
`
`Lobby, Inc., 477 U.S. 242, 248 (1986). A fact is “material” if it “might affect the
`
`outcome of the suit under the governing law.” Id.
`
`The moving party bears the initial burden of showing that there is no genuine
`
`18
`
`dispute of material fact and that it is entitled to prevail as a matter of law. Celotex, 477
`
`19
`
`U.S. at 323. If the moving party does not bear the ultimate burden of persuasion at trial,
`
`20
`
`it nevertheless “has both the initial burden of production and the ultimate burden of
`
`21
`
`22
`
`
`
`persuasion on a motion for summary judgment.” Nissan Fire & Marine Ins. Co. v. Fritz
`
`Companies, Inc., 210 F.3d 1099, 1102 (9th Cir. 2000). “In order to carry its burden of
`
`ORDER - 14
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 15 of 22
`
`
`
`production, the moving party must either produce evidence negating an essential element
`
`of the nonmoving party’s claim or defense or show that the nonmoving party does not
`
`have enough evidence of an essential element to carry its ultimate burden of persuasion at
`
`trial.” Id. If the moving party meets its burden of production, the burden then shifts to
`
`the nonmoving party to identify specific facts from which a factfinder could reasonably
`
`find in the nonmoving party’s favor. Celotex, 477 U.S. at 324; Anderson, 477 U.S. at
`
`250.
`
`The court is “required to view the facts and draw reasonable inferences in the light
`
`most favorable to the [nonmoving] party.” Scott v. Harris, 550 U.S. 372, 378 (2007).
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`
`The court may not weigh evidence or make credibility determinations in analyzing a
`
`11
`
`motion for summary judgment because these are “jury functions, not those of a judge.”
`
`12
`
`13
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`Anderson, 477 U.S. at 249-50. Nevertheless, the nonmoving party “must do more than
`
`simply show that there is some metaphysical doubt as to the material facts . . . . Where
`
`the record taken as a whole could not lead a rational trier of fact to find for the
`
`nonmoving party, there is no genuine issue for trial.” Scott, 550 U.S. at 380 (quoting
`
`Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 586-87 (1986) (internal
`
`quotation marks omitted)).
`
`B.
`
`Extraterritoriality Doctrine
`
`Under Illinois law, a “statute is without extraterritorial effect unless a clear intent
`
`20
`
`in this respect appears from the express provisions of the statute.” Avery v. State Farm
`
`21
`
`Mut. Ins. Co., 835 N.E.2d 801, 852 (Ill. 2005) (quoting Dur-Ite Co. v. Indus. Comm’n, 68
`
`22
`
`N.E.2d 717 (Ill. 1946) (internal quotation marks omitted)). Because BIPA does not
`
`
`
`ORDER - 15
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 16 of 22
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`contain such an express provision, it does not apply extraterritorially to conduct outside
`
`of Illinois. Rivera v. Google Inc., 238 F. Supp. 3d 1088, 1100 (N.D. Ill. 2017); (see
`
`3/15/21 Order at 6). Thus, to survive summary judgment, Plaintiffs must show a genuine
`
`issue of material fact regarding whether the circumstances underlying their BIPA claims
`
`“occurred primarily and substantially in Illinois.” Avery, 835 N.E.2d at 854; (see also
`
`3/15/21 Order at 6).
`
`Microsoft asserts that Illinois’s extraterritoriality doctrine bars Plaintiffs’ BIPA
`
`claims because none of its conduct relating to those claims took place in Illinois. (Mot. at
`
`11-15.) Rather, its relevant conduct—downloading, reviewing, and evaluating the DiF
`
`10
`
`Dataset—took place in Washington and New York. (Id. at 13-15.) Thus, according to
`
`11
`
`Microsoft, Plaintiffs cannot prove that its conduct “occurred primarily and substantially
`
`12
`
`in Illinois.” (Id. at 10-11 (citing Avery, 835 N.E.2d at 854).) Plaintiffs, for their part,
`
`13
`
`counter that the extraterritoriality doctrine does not apply because Microsoft’s relevant
`
`14
`
`conduct occurred in Illinois. (Resp. at 10-16.) The court agrees with Microsoft that the
`
`15
`
`extraterritoriality doctrine bars Plaintiffs’ BIPA claims as a matter of law.
`
`16
`
`Plaintiffs have not met their burden at summary judgment to establish a genuine
`
`17
`
`issue of material fact regarding whether Microsoft’s relevant conduct “occurred primarily
`
`18
`
`and substantially in Illinois.” Avery, 835 N.E.2d at 854. First, Plaintiffs rely on the
`
`19
`
`court’s order denying Microsoft’s motion to dismiss, in which the court identified the
`
`20
`
`allegations in Plaintiffs’ complaint that precluded dismissal on extraterritoriality grounds.
`
`21
`
`(Resp. at 13-14 (quoting 3/15/21 Order at 8).) At summary judgment, however, Plaintiffs
`
`22
`
`can no longer rest on their allegations. Instead, they must identify evidence sufficient to
`
`
`
`ORDER - 16
`
`
`
`Case 2:20-cv-01082-JLR Document 145 Filed 10/17/22 Page 17 of 22
`
`
`
`1
`
`2
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`establish a genuine issue of material fact regarding whether the circumstances giving rise
`
`to their claims occurred “primarily and substantially in Illinois.” Avery, 835 N.E.2d at
`
`854. As discussed below, they have not met this burden.
`
`Second, Plaintiffs contend that the extraterritoriality doctrine does not bar their
`
`claims because (1) Plaintiffs resided in Illinois; (2) Plaintiffs’ photos from which their
`
`biometric data was collected were taken in Illinois and uploaded to t