Long v. United States Immigration and Customs Enforcement

• Court: District Court, N.D. New York
• Decided: October 9, 2020
• Docket: 5:17-cv-00506
• Status: Unknown

Opinion

UNITED STATES DISTRICT COURT NORTHERN DISTRICT OF NEW YORK

SUSAN B. LONG and DAVID BURNHAM,

Plaintiffs, No. 5:17-cv-00506 (BKS/TWD)

v.

UNITED STATES IMMIGRATION AND CUSTOMS ENFORCEMENT,

Defendant.

Appearances: For Plaintiffs: Terence P. Keegan Zachary M. Press Miller Korzenik Sommers Rayman LLP 1501 Broadway, Suite 2015 New York, NY 10036 For Defendant: Antionette T. Bacon Acting United States Attorney Ransom P. Reynolds, III Assistant United States Attorney 100 South Clinton Street Syracuse, NY 13261

Ryan C. Stubbs U.S. Immigration and Customs Enforcement 500 12th St SW; MS 5900 Washington, DC 20536-5900 Hon. Brenda K. Sannes, United States District Judge: MEMORANDUM-DECISION AND ORDER I. INTRODUCTION Plaintiffs Susan B. Long and David Burnham brought this action under 5 U.S.C. § 552(a)(4)(B) against Defendant United States Immigration and Customs Enforcement (“ICE”) seeking to compel production of certain agency records pursuant to the Freedom of Information Act, 5 U.S.C. §§ 551–559 (“FOIA”). The Court denied the parties’ cross-motions for summary judgment on September 27, 2018. (Dkt. No. 32). ICE filed a supplemental declaration on April 29, 2019, (Dkt. No. 58-1), and moved for reconsideration of its motion for summary judgment. (Dkt. No. 58). Plaintiffs responded on May 16, 2019; citing to alleged contradictions and

inconsistencies in the information ICE has provided, Plaintiffs sought an evidentiary hearing. (Dkt. No. 59). The Court granted that request, and held an evidentiary hearing on August 15, 2019. (Dkt. No. 66). Both parties have submitted final letter briefs, and the Court heard oral argument on September 23, 2020. (Dkt. Nos. 90-92). Having carefully considered the testimony presented at the evidentiary hearing, as well as the parties’ arguments and all of the submissions in this case, the Court denies ICE’s renewed motion for summary judgment without prejudice to renewal. II. BACKGROUND1 Plaintiffs are the co-founders and co-directors of the Transactional Records Access Clearinghouse (“TRAC”), a data gathering, data research and data distribution center sponsored

by the Whitman School of Management and the Newhouse School at Syracuse University. (Dkt. No. 19-11, ¶ 73). TRAC gathers a significant amount of immigration enforcement data from ICE through FOIA requests, and makes the information available to the public in online databases. (Dkt. No. 19-11, ¶¶ 76-78). In the two FOIA requests at issue here, Plaintiffs seek information relating to ICE’s use of detainers and notices of release, i.e., ICE’s “requests to maintain custody of an individual for up to 48 hours beyond when he or she would have otherwise been released,”

1 The Court assumes familiarity with the procedural and factual history of this case, as set out in the Court’s September 27, 2018 memorandum decision and order. See Long v. ICE, No. 17-cv-0506, 2018 WL 4642824, at *1- 3, 2018 U.S. Dist. LEXIS 166170, at *1-10 (N.D.N.Y. Sept. 27, 2018). and ICE’s requests for notice from a law enforcement agency “of the pending release from custody of a suspected priority removable individual at least 48 hours prior to release, if possible.” (Dkt. No. 19-1, ¶ 9; Dkt. No. 15-1, ¶¶ 12-14). Plaintiffs seek “person-by-person anonymous data” relating to individuals subject to ICE detainers and notices, including their demographic characteristics, immigration history, and criminal background. (Dkt. No. 1-1, at 1;

Dkt. No. 1-5, at 1). Although Plaintiffs have been submitting FOIA requests for records concerning ICE’s use of I-247 forms2 since 2011, Plaintiffs allege that from January 2017 onward, ICE has refused to provide “much of the information produced in its previous responses.” (Dkt. No. 19-11, ¶¶ 77, 82). According to ICE, Plaintiffs’ FOIA request number 2017-ICFO-08061 “requested 150 total data points related to ICE’s use of I-247/I-247D detainers for ‘FY 2015 through November 2016,’” and Plaintiffs’ FOIA request number 2017-ICFO-08062 “requested 131 total data points related to ICE’s use of I-247N detainers for ‘November FY 2016.’” (Dkt. No. 58-1, ¶¶ 39, 49).3 ICE produced an Excel spreadsheet in response to each request, (Dkt. No. 19-11, ¶¶ 50, 65),

providing information that “existed in the database in association with detainers.” (Dkt. No. 58- 1, ¶¶ 40, 51). According to Operations Research Analyst Patricia de Castro, ICE produced 74

2 ICE used form I-247 for detainers prior to July 2015. (Dkt. No. 15-1, ¶ 12). Since July 2015, ICE has made two distinct forms available for use by its employees: detainer form I-247D, which is used to request that a law enforcement agency maintain custody of an individual for up to 48 hours, and detainer form I-247N, which is used to request that a law enforcement agency notify ICE of an individual’s pending release from custody. (Dkt. No. 19- 1, ¶ 9; Dkt. No. 15-1, ¶¶ 12-14). 3 In her initial declaration in support of ICE’s motion for summary judgment, Marla Jones, the Unit Chief of the Statistical Tracking Unit within Enforcement and Removal Operations Law Enforcement and System Analysis at ICE, provided a slightly different number of the total data points sought and produced with respect to each request. (See Dkt. No. 15-1, ¶¶ 19, 32 (stating that Plaintiffs’ FOIA request number 2017-ICFO-08061 “sought a total of 148 data points” and ICE produced 72 data points, and that Plaintiffs’ FOIA request number 2017-ICFO-08062 “contained a total of 132 total data points,” and ICE produced 46 data points)). The Court has used the numbers from ICE’s most recent declaration, from Patricia de Castro, (Dkt. No. 58-1), but the exact number is not determinative. data points for request number 2017-ICFO-8061 and 45 data points for request number 2017- ICFO-08062. (Dkt. No. 58-1, ¶¶ 40, 51). The data that ICE has not produced includes, inter alia, information about what happened following each detainer, including whether “the subject of a detainer was ever taken into ICE custody, whether that individual had ever been charged with or convicted of a crime, what any

such offenses were, and whether the individual was deported.” (Dkt. No. 25-1,¶ 7). Plaintiff Susan Long, the co-founder and co-director of TRAC, states that ICE’s failure to continue to provide such information “prevents the public from being informed about the extent of [ICE’s] use of detainers, the cooperation (or lack thereof) received from local and state law enforcement agencies, and the effectiveness of detainers in achieving this administration’s immigration enforcement goals.” (Dkt. No. 25-1, ¶ 10). A. ICE’s Motion for Summary Judgment In November 2017, ICE moved for summary judgment, arguing that it had conducted a line-by-line review of Plaintiffs’ requests and fully complied with its obligations under FOIA. (Dkt. No. 15-3).4 ICE stated that it did not produce certain data requested: (1) “because the

request in effect asked a question or constituted an implied question” rather than a request for a record; (2) the data points were not reasonably described or did not exist in the form requested without additional analyses or calculations; and (3) the data points simply did “not exist in the database as written.” (Dkt. No. 15-3, at 13-22). ICE argued that responding to a request that falls into one of these categories would impermissibly require it to create “calculations and new data

4 ICE submitted declarations from Marla Jones, the Unit Chief of the Statistical Tracking Unit within Enforcement and Removal Operations Law Enforcement and Systems Analysis at ICE, and Catrina Pavlik-Keenan, the FOIA Officer at ICE’s Freedom of Information Act Office, in support of its motion for summary judgment. (Dkt. Nos. 15- 1, 15-2). search and query methodologies[,] as they relate to data that ICE does not track in the manner in which the data is requested.” (Dkt. No. 23-1, ¶ 13). Plaintiffs moved for summary judgment as well, arguing that ICE’s declarations were inadequate to support its motion for summary judgment because they did not contain reasonable specificity of detail, were called into question by contradictory evidence in the record, and

misconstrued certain of Plaintiffs’ requests. (Dkt. No. 19-12, at 11-17; Dkt. No. 25, at 9-12). Plaintiffs argued, among other things, that many of ICE’s assertions were contradicted by data that ICE had produced in response to FOIA requests which are the subject of another case, Long v. ICE, No. 17-cv-01097 (D.D.C.) (“the 2017 D.C. Action”).5 This Court denied both motions for summary judgment, concluding that Plaintiffs met their burden of showing that summary judgment was not appropriate and citing to ICE’s misconstruction of certain requests, as well as Plaintiffs’ production of “tangible evidence regarding ICE’s response to other FOIA requests and inconsistencies within ICE’s declarations.” Long v. ICE, No. 17-cv-0506, 2018 WL 4642824, at *8, 2018 U.S. Dist. LEXIS 166170, at *23-

24 (N.D.N.Y. Sept. 27, 2018). B. ICE’s Renewed Motion for Summary Judgment ICE renewed its motion for summary judgment based on the Supplemental Declaration of Patricia de Castro (the “de Castro Declaration”), which primarily contends that: (i) ICE’s search for records was adequate, because it has already provided all of the data that exists in its database in association with its use of I-247/I-247D and I-247N detainers, (Dkt. No. 58-1, ¶¶ 39-41, 49- 51); (ii) additional data related to individuals for whom detainers have been issued is not subject

5 In the 2017 D.C. Action, the Plaintiffs seek information regarding removals, i.e., “anonymous case-by-case information about each person whom ICE deported as a result of” certain immigration enforcement programs. Long v. ICE, 2018 WL 4680278, at *1, 2018 U.S. Dist. LEXIS 167222, at *3 (D.D.C. Sept. 28, 2018). to disclosure under FOIA because it either does not exist with respect to the detainers population in its database, (id. ¶¶ 47-48, 55), or would “require analysis, calculations, and the creation of new records,” (id. ¶¶ 42, 52); and (iii) continuing to accommodate Plaintiffs’ requests for data beyond the information ICE has already agreed to provide would be unduly burdensome, (id. ¶¶ 20-24). Plaintiffs sought an evidentiary hearing to “resolve the new contradictions in ICE’s

testimony” they claimed the de Castro Declaration created, including those regarding ICE’s database search capabilities, the time required to respond to Plaintiffs’ requests, and whether the requested records exist within ICE’s database. (Dkt. No. 59). The Court granted that request, and directed ICE to have present at the hearing the official(s) necessary to address the concerns raised by Plaintiff and the “inconsistencies in ICE’s responses to FOIA requests.” (Dkt. No. 64). C. The Evidentiary Hearing At the August 15, 2019 evidentiary hearing, ICE presented the testimony of Curtis Hemphill, a detention and deportation officer within ICE’s Enforcement and Removal Operations Statistical Tracking Unit (“STU”). (Dkt. No. 80, at 8-9). Among his other

responsibilities at STU, Hemphill reviews incoming FOIA requests, assists ICE analysts and statisticians with determining “which data points we will pull and provide in response to those requests,” and clears ICE’s FOIA responses for release. (Id. at 10). Plaintiffs called Patricia de Castro, the declarant on ICE’s renewed motion for summary judgment. (Dkt. No. 58-1; Dkt. No. 80, at 79). As an Operations Research Analyst in ICE’s Enforcement and Removal Operations, Law Enforcement Systems Analysis, Data Driven Management Unit, De Castro works with “a team of analysts, statisticians, detention and deportation officers, program analysts, and mission support staff to support data-driven management and FOIA processes and litigation.” (Dkt. No. 58-1, ¶ 1). Plaintiffs also called Plaintiff Susan B. Long. (Dkt. No. 80, at 111). Plaintiffs sought to continue the evidentiary hearing to present expert testimony of Dr. Paul Clark, and ICE opposed that request. (Dkt. Nos. 73, 84-86).6 ICE challenged Dr. Clark’s expert qualifications “based upon the fact that he has never seen, used, or reported data for the FOIA or any ICE operational reporting need, from ICE’s database.” (Dkt. No. 85, at 3). After considering the parties’ letter briefing and a declaration of Dr. Clark, the Court denied Plaintiffs’

motion to continue the hearing for the testimony of Dr. Clark, but stated that it would consider Dr. Clark’s declaration, and permitted ICE to submit a declaration in response. (Dkt. No. 87). Following the Court’s ruling, Plaintiffs and ICE submitted final letter briefs,7 and the Court heard oral argument on the issues of whether ICE’s search was adequate and whether fulfilling Plaintiffs’ requests would require the creation of new records. (Dkt. Nos. 90-92). The following information regarding the ICE databases and the work necessary to respond to the FOIA requests at issue is drawn from the testimony presented at the evidentiary hearing and the declarations submitted by ICE officials in this case. 1. The EID and the IIDS Databases ICE maintains two databases that are relevant here: the Enforcement Integrated Database

(“EID”) and the ICE Integrated Decision Support System (“IIDS”). The EID is ICE’s “common database repository for all records created, updated, and accessed by a number of software applications.” (Dkt. No. 15-1, ¶ 6). The EID “captures and maintains information related to the

6 One week before the August 15, 2019 evidentiary hearing, Plaintiffs submitted a letter seeking “an opportunity to continue the evidentiary hearing at a later date, if needed” because they “may need to rely on the expert testimony of Dr. Paul C. Clark to rebut the August 15 testimony from Defendant’s witnesses.” (Dkt. No. 73). When Plaintiffs sought, near the close of the evidentiary hearing, to continue the hearing to present the testimony of Dr. Clark, Defendant objected to a continuance and questioned the basis of any testimony by Dr. Clark. (Dkt. No. 80, at 102- 08). Although the Court stated that it was “not sure how Dr. Clark has anything” relevant to ICE’s argument that responding to the FOIA requests would require ICE to create a record, and “[t]o the extent” Dr. Clark “had information” on that issue, that information “could have been used for cross-examining the witnesses,” but nothing “that appeared to be of that nature,” had been used during cross-examination, the Court permitted Plaintiffs to make a letter proffer of Dr. Clark’s proposed testimony. (Dkt. No. 80, at 109-10). 7 ICE opted not to submit an additional declaration rebutting Dr. Clark’s testimony. (Dkt. No. 90). investigation, arrest, booking, detention, and removal of persons encountered during immigration and law enforcement investigations and operations conducted by ICE and [United States Customs and Border Protection (“CBP”)].” (Id. ¶ 7). It is a live database that is “used to store data captured 24 hours a day” by law enforcement personnel from ICE and CBP. (Id. ¶ 8). Users of the EID can access a “person-centric and/or event-centric view” of EID data through “a

number of software applications,” and “print records containing the EID data” but, according to de Castro, the EID does not have a reporting capability. (Dkt. No. 15-1, ¶¶ 6-7; Dkt. No. 58-1, ¶ 19).8 To respond to Plaintiffs’ FOIA requests, ICE searched the IIDS, and did not search the EID. (Dkt. No. 15-1, ¶ 9). “ICE queries IIDS instead of the EID to protect the integrity of the live data held in the EID operational environment and to prevent the performance of EID from being diminished.” (Id. at ¶ 11). The IIDS is a separate database that is a “snapshot” of “a subset of data from” the EID. (Dkt. No. 80, at 12; see Dkt. No. 15-1, at 3). The IIDS is “the primary source of reporting and responding to FOIA requests for the STU.” (Dkt. No. 80, at 12). The IIDS

database is updated with data from the EID three days a week during an Extract, Transform, Load (“ETL”) cycle. (Dkt. No. 15-1, ¶ 28 n.1). During the ETL cycle, data from the EID is “optimized and reformatted in a way that aids reporting” before being “added into the IIDS reporting database.” (Id.). The IIDS database was “designed to allow [ICE] to prepare and publish regular summary reports regarding law enforcement actions.” (Dkt. No. 58-1, ¶ 26).

8 In her initial declaration in support of ICE’s motion for summary judgment, Marla Jones stated that “[t]he EID allows ICE officers to manage cases from the time of an alien’s arrest, in-processing, or placement into removal proceedings, through the final case disposition.” (Dkt. No. 15-1, ¶ 6). At the evidentiary hearing, Curtis Hemphill testified that this is not accurate. (Dkt. No. 80, at 23). Hemphill explained that “case management is not done in the EID”; case management is done in an up-front application called the ENFORCE Alien Removal Module” (“EARM”). (Id.). “[C]ases are managed in EARM using data that is brought into EARM from the EID.” (Id. at 47- 48). On a weekly basis, ICE runs a series of predefined queries in IIDS to generate specific “populations” of data that correspond to particular types of law enforcement actions, including a detainers population, a removals population and an arrests population. (Dkt. No. 80, at 12, 39). These queries “are created in the course of [ICE’s] regular business for the purpose of . . . meeting [its] regular reporting obligations.” (Id. at 39). Hemphill described a population as “a

grouping of records that are like in scope” such as “detainers, arrests, encounters, [and] removals,” which exist “in a grouping outside of the database.” (Dkt. No. 80, at 43-44). ICE uses the populations, rather than the IIDS database as a whole, “to respond to FOIA requests and to create reports.” (Dkt. No. 80, at 12). Thus, in responding to FOIA requests, ICE searches particular populations that correspond to the category of information sought. For example, when ICE receives a FOIA request related to detainers, it searches the detainers population for relevant data; similarly, when ICE receives a FOIA request related to removals, it searches the removals population for relevant data. (Id. at 12-13). Because each population is designed to capture data related to a different type of law enforcement event, and different data

points are entered into the system when different types of law enforcement events occur, the populations necessarily contain different data points. For example, the detainers population “include[s] the points that are captured on the [I-247/I-247D] form as well as some additional points” that are entered at the time a detainer is recorded; similarly, there are a “set number of removals data points” that are recorded at the time a removal occurs, and are found in the removals population but not the detainers population. (Id. at 13). Hemphill testified that ICE “files hundreds of thousands of detainers a year, so a resulting dataset from th[e] type of detainers request [in this case] is going to have approximately 70 columns and several hundred thousand rows.” (Id. at 67). He further testified that a detainer spreadsheet produced for TRAC’s requests would be approximately “63,000 pages” if printed. (Id. at 61). 2. Tracking an Individual Alien One key dispute between the parties revolves around the fact that Plaintiffs seek “person- centric” data—in other words, data that would allow them to view an individual’s complete

record of law enforcement actions—while the data used for reporting in the IIDS “is not arranged by individuals, but by law enforcement actions (i.e., detainers, removals).” (Dkt. No. 58-1, ¶ 28). ICE avers that, while the information in the IIDS “is granular and detailed,” because the information is maintained “in distinct modules (law enforcement actions), i.e., detention, detainer, removal—versus tracking individuals,” the data it is “not neatly arranged to track individuals and their enforcement history, such as all detainers, all removals, etc.” (Dkt. No. 58- 1, ¶ 25). Importantly, ICE states that there is not “one single identifier for the same ‘alien’ that would then be sequential in nature and carry across each [ETL] cycle change.” (Dkt. No. 15-1, ¶

28). Put another way, there is no single identifier for a particular individual that allows ICE to readily link data corresponding to that individual that exists in different populations (i.e. the “detainers” population” and the “removals” population). (Dkt. No. 80, at 53 (Hemphill’s testimony that “there are no identifiers that we have that I am aware of that exist in every record”)). When questioned why different law enforcement actions corresponding to the same individual, such as a detainer and a removal, could not be linked by any identifying number, Hemphill testified that “there are so many identifiers in our database that are generated based on the encounters and the events and the different law enforcement actions,” and that there is a risk of underreported data if only one identifier, such as a subject ID or a case ID or an “A number” is used to attempt to link the populations. (Dkt. No. 80, at 58-59).9 As a result, in the data that ICE uses for reporting, which exists in populations, “there is no function in [the] IIDS database to track someone from an individual perspective.” (Dkt. No. 80, at 17; see also Dkt. No 58-1, ¶ 26 (“[T]here is not one comprehensive master record that allows [analysts] to pull up all enforcement history (i.e. all arrests or book-ins) related to that individual”)).10

De Castro summarizes the problem as follows: “Ultimately, the problem is that Plaintiffs seek data (and extensive data connections) concerning individuals, and the ICE database structure does not accommodate that format.” (Id. ¶ 35). For example, while IIDS may contain data related to a detainer issued for a specific individual and data related to that individual’s subsequent arrest, that data exists in separate populations—the “arrest [data] is populated in the system but it’s not tied to a specific detainer.” (Dkt. No. 80, at 17). And because, as explained above, there is no one single identifier for an individual that exists across populations, ICE states that there is no way to readily link data in separate populations to the same individual without conducting additional analysis to reconcile the differing identifiers, thus “creat[ing] new data

connections that did not exist previously.” (Dkt. No. 58-1, at ¶¶ 42, 52). ICE acknowledges that it is able to “connect different law enforcement actions to individuals . . . when it’s operationally required”; however, it contends that doing so requires additional steps of analysis that result in the creation of new records, steps that ICE maintains is it is not required to undertake in responding to FOIA requests. (Dkt. No. 80, at 20). As De Castro explains, in order to follow an alien through the immigration law enforcement system by

9 In this case ICE provided several different identifiers in response to Plaintiffs’ FOIA requests: “Subject ID, EID Civ Pers ID, Detainer ID, and Eid Civ ID.” (Dkt. No. 58-1, at 35; Dkt. No. 59, at 2). 10 ICE states that the only “comprehensive master record of an individual” is the individual’s Alien File, or “A-File.” (Dkt. No. 80, at 27). A-Files are “the official paper file for all immigration and naturalization records created or consolidated since April 1, 1944, and are under the purview of U.S. Citizenship and Immigration Services,” not ICE. (Dkt. No. 58-1, ¶ 37). connecting data from different populations that corresponds to the same individual, ICE analysts would have to either: (1) “create new data combinations through extensive programming, analysis, calculations, and the creation of new reports” or (2) “create and maintain new randomized unique IDs after creating data connections between events for individuals.” (Dkt. No. 58-1, ¶ 12). Thus, while ICE will create a query to provide “data points associated with [the]

detainer[s] population” in response to “a request for detainers” and “data points associated with [the] removal [population]” in response to “a request for removals,” it will “not connect them” when responding to a FOIA request, as ICE’s position is that “pull[ing] . . . data points from disparate populations that are not related to each other” in the database requires creation of a “new record that did not exist prior to the request from [a] requester.” (Dkt. No. 80, at 37-38). Hemphill’s discussion of an application that law enforcement officers use in the field, the Enforce Alien Removal Module (EARM), was illustrative. The EARM is a case management application that directly utilizes data that is stored in the EID (as distinct from the IIDS). (Id. at 23). An EARM user can “create a case” in the system using an identifier such as an alien

number, an FBI number or a subject ID; creating a case generates a case ID that retrieves data from the EID linked to that person. (Id. at 24-25, 55-56). To link law enforcement actions such as detainers and removals, analysts would have to write queries that would do this; when ICE was previously providing TRAC the information at issue here, analysts wrote such queries. (Id. at 56). Given these features of the relevant databases, ICE states that certain data fields were produced in response to the FOIA requests at issue in the 2017 D.C. Action (concerning removals) that have not been produced in response to the FOIA requests at issue in this action (concerning detainers) because “[d]ifferent data is collected and entered when different law enforcement actions occur,” and that therefore, the data Plaintiffs received in response to their removals-related requests (which came from the removals population) and the data they received in response to their detainers-related requests (which came from the detainers population) necessarily includes different fields. (Dkt. No. 58-1, ¶ 13). As Hemphill explained, Plaintiffs ”get the entire population of detainers . . . removals and arrests and variations thereof. They just don’t

get them linked together because that would be creating a record that doesn’t exist.” (Dkt. No. 80, at 29). 3. Change in Policy De Castro explained that, “[f]rom 2011 to July 2016, ERO analysts discretionarily performed analysis, research and creation of new records to fulfill Plaintiffs’ FOIA requests,” but that as of July 2016, ICE is no longer willing to do so. (Dkt. No. 58-1, ¶ 11). De Castro stated that this change in policy regarding the records produced to Plaintiffs occurred on July 29, 2016, after consultation with the ICE Office of the Principal Legal Advisor. (Id. ¶ 15). De Castro stated that the change was due to the “increase in the overall volume and complexity of FOIA requests

submitted to ERO in the past two years” and the lack of “personnel and computational resources to perform beyond the requirements of FOIA by creating new records,” and was “based solely on the FOIA statute and the FOIA caselaw,” which does not require an agency to create records. (Id. ¶¶ 14-15). III. STANDARD OF REVIEW “FOIA cases are generally and most appropriately resolved on motions for summary judgment.” Families for Freedom v. U.S. Customs & Border Prot., 797 F. Supp. 2d 375, 385 (S.D.N.Y. 2011). To prevail on summary judgment, ICE bears the burden of establishing the adequacy of its search. Brennan Ctr. for Justice at New York Univ. Sch. of Law v. Dep’t. of Homeland Sec., 331 F. Supp. 3d 74, 84 (S.D.N.Y. 2018); see 5 U.S.C. § 552(4)(B) (“[T]he burden is on the agency to sustain its action.”). A search is “adequate” when it is “reasonably calculated to uncover all relevant documents.” Oglesby v. U.S. Dep’t of the Army, 920 F.2d 57, 68 (D.C. Cir. 1990). “The adequacy of a search is not measured by its results, but rather by its method,” and therefore, “a

search is not inadequate merely because it does not identify all responsive records.” N.Y. Times Co. v. U.S. Dep’t of Justice, 756 F.3d 100, 124 (2d Cir. 2014), opinion amended on denial of reh’g, 758 F.3d 436 (2d Cir. 2014), supplemented, 762 F.3d 233 (2d Cir. 2014). “[A]n agency’s search need not be perfect, but rather need only be reasonable.” Grand Cent. P’ship, Inc. v. Cuomo, 166 F.3d 473, 489 (2d Cir. 1999). “In resolving summary judgment motions in a FOIA case, a district court proceeds primarily by affidavits in lieu of other documentary or testimonial evidence.” Long v. Office of Pers. Mgmt., 692 F.3d 185, 190 (2d Cir. 2012). To show that its search was adequate, an agency must provide “‘a reasonably detailed affidavit, setting forth the search terms and the type of

search performed, and averring that all files likely to contain responsive materials . . . were searched.’” Roman v. Dep’t of Air Force, 952 F. Supp. 2d 166, 171 (D.D.C. 2013) (quoting Iturralde v. Comptroller of the Currency, 315 F.3d 311, 313-14 (D.C. Cir. 2003)). Affidavits or declarations “submitted by an agency are ‘accorded a presumption of good faith,’” and factual discovery “‘relating to [an] agency’s search . . . generally is unnecessary if the agency’s submissions are adequate on their face.’” Grand Cent. P’ship, Inc., 166 F.3d at 488-89 (alteration in original) (quoting Carney v. U.S. Dep’t of Justice, 19 F.3d 807, 812 (2d Cir. 1994)). Once an agency has satisfied its burden, the affidavits “cannot be rebutted” by a plaintiff’s “purely speculative claims about the existence and discoverability of other documents.” SafeCard Servs., Inc. v. S.E.C., 926 F.2d 1197, 1200 (D.C. Cir. 1991) (citation and internal quotation marks omitted). Rather, a plaintiff “must make a showing of bad faith on the part of the agency sufficient to impugn the agency’s affidavits or declarations, or provide some tangible evidence that . . . summary judgment is otherwise inappropriate.” Carney, 19 F.3d at 812 (citations omitted). Accordingly, summary judgment may be granted in favor of the agency

so long as its affidavits “contain reasonable specificity of detail rather than merely conclusory statements, and if they are not called into question by contradictory evidence in the record[,] or by evidence of agency bad faith.” Grand Cent. P’Ship., 166 F.3d at 478 (quoting Gallant v. NLRB, 26 F.3d 168, 171 (D.C. Cir. 1994)). While “evidentiary hearings are rare in the FOIA context,” where facts are genuinely disputed, “the agency’s burden requires it to prove by a preponderance of the evidence that the factual dispute should be resolved in its favor.” Long v. ICE, No. 14-cv-109, 2020 WL 2849904, at *5, 2020 U.S. Dist. LEXIS 96468, at *15 (D.D.C. June 2, 2020). IV. DISCUSSION The 1996 Electronic FOIA Amendments (“E-FOIA Amendments”) “altered FOIA’s

definition of ‘search’ and agency ‘record’ to specifically address automatic review of electronic files.” Everytown for Gun Safety Support Fund v. Bureau of Alcohol, Tobacco, Firearms & Explosives, 403 F. Supp. 3d 343, 348 (S.D.N.Y. 2019); see Pub. L. 104-231, 110 Stat. 3048 (1996); 1996 U.S.C.C.A.N. 3448. An agency responding to a FOIA request must “make reasonable efforts to search” for “records in electronic form or format, except when such efforts would significantly interfere with the operation of the agency’s automated information system.” 5 U.S.C. § 552(a)(3)(C). A “search” “means to review, manually or by automated means, agency records for the purpose of locating those records which are responsive to a request.” 5 U.S.C. § 552(a)(3)(D). And “records” include electronically stored information. 5 U.S.C. § 552(f)(2)(A). Here, ICE argues that it is not obligated to respond to the requests at issue because doing so would require it to go beyond the foregoing obligations by creating new records and conducting research and analysis.11 It is well-settled that FOIA “does not obligate agencies to create documents or retain documents; it only obligates them to provide access to those which it in fact has created and retained.” Kissinger v. Reporters Committee for Freedom of the Press,

445 U.S. 136, 152 (1980); see also American Civil Liberties Union v. Dep’t of Justice, 681 F.3d 61, 71 (2d Cir. 2012). Nor does FOIA require an agency to conduct research beyond conducting a reasonable search for the documents sought in the FOIA request. McLean ex rel. J.N.M. v. Social Sec. Admin., No. 17-cv-06978, 2019 WL 1074273, at *5, 2019 U.S. Dist. LEXIS 1170, at *14-15 (S.D.N.Y. March 6, 2019); Nat’l Sec. Counselors v. CIA (“Nat’l Sec. Counselors II”), 960 F. Supp. 2d 101, 160 n.28 (D.D.C. 2013); Kissinger, 445 U.S. at 159 n.2 (Brennan, J., concurring in part). “Agencies are not required to ‘dig out all the information that might exist, in whatever form or place it might be found, and to create a document that answers plaintiff’s questions.’” Tokar v. U.S. Dep’t of Justice, 304 F. Supp. 3d 81, 91 (D.D.C. 2018) (quoting Frank

v. U.S. Dep’t of Justice, 941 F. Supp. 4, 5 (D.D.C. 1996)). Thus, “[a]fter the E-FOIA Amendments, whether information in a database constitutes an agency record hinges not on whether the information is housed in the form requested, but whether generating the information requires the agency to engage in additional research or conduct additional analyses above and beyond the contents of its database.” Everytown for Gun Safety Support Fund, 403 F. Supp. 3d at 359. On this question, the parties agree that ICE is obligated to sort “a pre-existing database of information to make information intelligible.” Id. at

11 ICE has not renewed the other arguments it made in its initial motion for summary judgment, and the Court has therefore not considered those arguments here. 356 (quoting Nat’l Sec. Counselors v. CIA (“Nat’l Sec. Counselors I”), 898 F. Supp. 2d 233, 270 (D.D.C. 2012)); (Dkt. No. 15-3, at 10; Dkt. No. 91, at 6). As ICE acknowledges, “[s]orting a database by a particular data field (e.g., date, category, title) is essentially ‘the application of codes or some form of programming,’ and thus does not involve creating new records or conducting research—it is just another form of searching that is within the scope of an agency’s

duties in responding to FOIA requests.” (Dkt. No. 15-3, at 10 (quoting 1996 U.S.S.C.A.N. 3448, 3465, House Report No. 104-795 on the E-FOIA Amendments)). And, as ICE further acknowledges, “extracting and compiling data does not amount to the creation of a new record as to any discrete pieces of information that the agency does possess in its databases and which are sought by Plaintiffs.” (Dkt. No. 15-3, at 10-11). Thus, while a search to extract and compile data within a pre-existing database is required under FOIA, a search requiring the agency to perform research or create a new record is not. Both parties cite the Department of Homeland Security regulation concerning electronic records and searches under FOIA, which serves to further elucidate the distinction between

“extracting and compiling” data and “creating a record.” 6 C.F.R. § 5.4(i). Under this regulation “[c]reating a computer program that produces specific requested fields or records contained within a well-defined database structure usually is considered business as usual.” 6 C.F.R. § 5.4(i)(2)(ii). On the other hand, “creating a computer program to merge files with disparate data formats and extract specific elements from the resultant file,” is “not considered business as usual, but a special service” which the agency is not required to perform. Id. There is no dispute here that ICE has the capacity to provide the information Plaintiffs seek. However, FOIA “only requires disclosure of certain documents which the law requires the agency to prepare or which the agency has decided for its own reasons to create.” NLRB v. Sears, Roebuck & Co., 421 U.S. 132, 162 (1975). “In responding to a FOIA request for ‘aggregate data’ . . . an agency need not create a new database or . . . reorganize its method of archiving data.” Nat’l Sec. Counselors I, 898 F. Supp. 2d at 270. Therefore, the fact that ICE is “capable of creating the record” is “insufficient to require the agency to produce such a record” if the record is “not one that [ICE] has in fact chosen to create and retain.” Nat’l Sec. Counselors II, 960 F.

Supp. 2d at 160 (emphasis in original) (quoting Yeager v. Drug Enf’t Admin., 678 F.2d 315, 321 (D.C. Cir. 1982)). Relying on these principles, ICE asserts that it is not obligated to produce data that requires following an alien through the system because ICE does not maintain that data in its system, in the form requested by Plaintiffs, and producing that information would require ICE to conduct research and create a new record. In other words, while discrete pieces of information might be stored within IIDS as a “granular” record, the data are not arranged to track individuals and their enforcement history as Plaintiffs request. Instead, ICE has “chosen to create and retain” data points in populations related to specific law enforcement actions. According to ICE,

complying with the requests here would involve more than writing a query to extract discrete data or fields from a database; responding to the requests would require a query that analyzes data from different populations to link data that is not linked in the database, essentially creating a new, comprehensive alien record that did not previously exist. ICE thus argues that this is unlike the situation where an agency is asked to extract and compile specific records that an agency acknowledges is stored in its database, see, e.g., Schladetsch v. Housing & Urban Dev., No. 99-0175, 2000 WL 33372125, at *3, 2000 U.S. Dist. LEXIS 22895, at *9 (D.D.C. Apr. 4, 2000), or a case in which the data requested is in a searchable field, Everytown for Gun Safety Support Fund, 403 F. Supp. 3d at 357. Here, according to ICE, Plaintiffs’ request for data tracking aliens across different populations requires an analysis of differing identifiers within distinct data sets, and the creation of new records produced from that analysis. At oral argument, when asked to analogize this to a paper records system, ICE argued that if ICE maintained records of detainers in drawers and subfolders and records of removals in other drawers with subfolders, FOIA would not require ICE to produce

the records Plaintiffs seek, linking the alien’s records that are stored in the two separate drawers. A. Adequacy of the Search Plaintiffs assert that ICE’s search of its detainer “population” from the IIDS was not an adequate search because ICE has failed to show that the search of this limited set of data satisfied ICE’s burden to search “all files likely to contain responsive materials.” (Dkt No. 91, at 3, 7-10). ICE contends that the IIDS, which “is the primary source of information that STU uses for reporting and responding to FOIA requests,” constitutes “the only responsive ICE database” that ICE was required to search, and that it has produced all responsive data from the IIDS’s detainers population that can be produced without, in ICE’s view, “creating a new record.” (Dkt. No. 90, at 3-5).

On the record currently before the Court, ICE has not met its burden of proving that its search in response to Plaintiffs’ FOIA requests was “adequate” within the meaning of the FOIA case law. To meet this burden, an agency need not “search every record system” that it has access to, but it also “cannot limit its search to only one record system if there are others that are likely to turn up the information requested,” Oglesby, 920 F.2d at 68, and it must satisfy the Court that it searched “all files likely to contain responsive materials.” Roman, 952 F. Supp. 2d at 171 (quoting Iturralde, 315 F.3d at 313-14) (emphasis added). Thus, to justify limiting its search to one particular record source, the agency must, “[a]t the very least” “explain in its affidavit that no other record system was likely to produce responsive documents.” Oglesby, 920 F.2d at 68. Here, Plaintiffs contend that FOIA requires ICE to search “databases – not [] ‘populations,’” and that ICE’s decision to limit its search to the detainers population— a subset of the IIDS database that is generated on a weekly basis through a pre-set query, and includes

only data points recorded by ICE at the time a detainer is filed, (Dkt. No. 80, at 12-13, 44-46)— failed to satisfy its obligation to conduct an adequate search. (Dkt. No. 91, at 1, 7-9). The Court finds that ICE has failed to meet its burden to explain why limiting its search to the detainers population was adequate. It is undisputed that additional data responsive to Plaintiffs’ requests exists in the IIDS outside of the detainers population, including in other populations, and prior to July 2016, ICE regularly produced this data to Plaintiffs. (Dkt. No. 58-1, ¶ 11; Dkt. No. 19-11, ¶¶ 77, 82; Dkt. No. 80, at 56). Notwithstanding that it willingly produced this data prior to July 2016, ICE maintains that it is not required under FOIA to continue to produce this data because the analysis required to “link” data from disparate populations constitutes the creation of a new

record. This contention is addressed separately below, but even assuming arguendo that ICE is correct that this analysis would constitute record creation, ICE has not explained whether, as an alternative to searching for data within specific, pre-generated populations and then conducting analysis to link that data to data in other populations, it is possible to simply query the IIDS as a whole for responsive data at the outset.12 (Dkt. No. 84-1, ¶ 9(c), (e), (h)). Assuming such a query is possible, ICE has not explained why such a query of the IIDS database would constitute more than “[c]reating a computer program that produces specific

12 Indeed, when directly asked this question at oral argument, counsel to ICE admitted that he did not know whether doing so was possible, and none of the other evidence ICE has produced addresses this point. requested fields or records contained within a well-defined database structure,” which the regulations make clear is “business as usual,” 6 C.F.R. § 5.4(i)(2)(ii); or, put another way, why such a query would not merely be the type of “extracting and compiling data” that ICE agrees does not amount to the creation of a record. (Dkt. No. 15-3, at 10-11); see also Long v. CIA, No. 15-cv-1734, 2019 WL 4277362, at *4, 2019 U.S. Dist. LEXIS 153618, at *11-12 (D.D.C. Sept.

10, 2019) (ruling that compiling records of “certain data fields that the CIA concedes exist within the database” did “not constitute the creation of a new record”); Schladetsch v. Housing & Urban Dev., No. 99-0175, 2000 WL 33372125, at *3, 2000 U.S. Dist. LEXIS 22895, at *9 (D.D.C. Apr. 4, 2000) (noting that “the programming necessary” to extract and compile “the discrete pieces of information” in the agency’s database “is a search tool and not the creation of a new record”). Thus, ICE has failed to explain how its search of the detainers population constituted a search of all files likely to contain responsive materials. Plaintiffs also question ICE’s failure to search the EID, noting that ICE has produced data from the EID in the past; acknowledged in another FOIA action that “person-centric unique

identifiers are reasonably likely to be found in EID”; and can retrieve data from the EID through its EARM system. (Dkt. No. 91, at 8-9 (citing American Immigration Council v. ICE, No. 18- 1614, 2020 WL 2748515, at *17, 2020 U.S. Dist. LEXIS 92794, at *25-26 (D.D.C. May 27, 2020)).13 ICE’s declarants acknowledge that the IIDS includes only a “subset” of data from the EID, (Dkt. No. 15-1, ¶ 9), and that person-centric data of the type Plaintiffs seek may be viewed and printed directly from the EID. (Id. ¶¶ 6-7). While ICE states that the EID does not have “a

13 In their first memorandum of law in this case, Plaintiffs noted that the EID is “the relevant database for [Plaintiffs’] requested records,” but concluded that the differences between the EID and IIDS “do not affect Plaintiffs’ current requests for relief” “because ICE has not represented that it has withheld data in the EID if it does not exist in the IIDS, or that it could do so.” (Dkt. No. 19-12, at 9 n.4). In their latest submission Plaintiffs note that at the time of their initial brief, “any distinctions between the EID and IIDS did not seem to affect Plaintiffs’ requests for relief” because ICE had asserted that the records did not exist in either database. (Dkt. No. 91, at 2). reporting capability,” (Dkt. No. 58-1, ¶ 19), and that it did not search the EID in order to “protect the integrity of the live data held in the EID operational environment and to prevent the performance of EID from being diminished,” (Dkt. No. 15-1, ¶ 11), it is not clear how these conclusory explanations square with the questions raised by Plaintiffs, and whether there may be responsive data within the EID that is not in the IIDS. And as Plaintiffs point out, at least one

Court reviewing another FOIA dispute has found that ICE failed to meet its burden to demonstrate an “adequate search” where it only searched the IIDS and not the EID. See American Immigration Council, 2020 WL 2748515, at *7-9, 2020 U.S. Dist. LEXIS 92794, at *20-26 (finding a genuine dispute of fact over adequacy of ICE’s search for records in response to plaintiffs’ request for person-centric “unique identifiers” where ICE only searched the IIDS and not the EID, including because ICE’s “own declarant acknowledges that person-centric unique identifiers are reasonably likely to be found in EID”). Here, because of the lack of factual development regarding ICE’s capability of searching its databases for the type of records Plaintiffs seek, the Court cannot conclude that ICE has met

its burden of proving its search was adequate. B. Creation of a Record Plaintiffs also assert that ICE has failed to meet its burden of showing that a search for the requested information would constitute creating a new record. Plaintiffs note that ICE has never particularized the “‘analysis and calculations’ required for the data points it has withheld.” (Dkt. No. 91, at 10). Relying on Hemphill’s testimony and the de Castro Declaration, ICE reiterates that linking data from disparate populations within the IIDS requires the “creation of complex queries which combine disparate data populations” and “create newly created data that did not exist prior to the receipt of the FOIA requests.” (Dkt. No. 90, at 6). As an initial matter, the Court recognizes that “[a]ffidavits submitted by an agency are accorded a presumption of good faith,” Carney, 19 F.3d at 812, and that “[a]ny factual assertions contained in [an agency’s] affidavits and other attachments in support of motions for summary judgment are accepted as true unless the nonmoving party submits affidavits or other documentary evidence contradicting those assertions.” Roman, 952 F. Supp. 2d at 171 (quoting

Wilson v. Dep’t of Transp., 730 F. Supp. 2d 140, 148 (D.D.C. 2010)). At the same time, where, as here, the process of compiling the records Plaintiffs seek involves “the application of codes or some form of programming to retrieve the information” from an electronic database, it is the Court’s task to determine whether, as a legal matter, the process described by the agency “crosses the all-important line between searching a database, on the one hand, and either creating a record or conducting research in a database on the other.” Nat’l Sec. Counselors I, 898 F. Supp. 2d at 270-71. As such, as a “general rule,” the facts asserted by the agency in support of its conclusion that it did not need to produce the requested information must be sufficiently “detailed” and “non-conclusory” to allow the Court to determine whether the agency’s position

is legally justified. Id. at 250. As explained above, the de Castro Declaration and Hemphill’s testimony explain in general terms that the data points Plaintiffs seek are contained in separate populations and that creating links between populations requires additional analysis. See supra Section II.C.1-2. However, neither de Castro nor Hemphill, nor any of ICE’s other declarants, has explained what additional “analysis,” “calculations” or “research” is required to create such links, or what specific steps are involved. (Dkt. No. 58-1, ¶¶ 42, 52; see also Dkt. No. 80, at 58 (Hemphill’s testimony, when directly asked how analysts create links between data in the removal and detainer populations, that “[t]hat’s way beyond the scope of my . . . technical abilities”); Id. at 101-102 (de Castro’s testimony that she could not explain with specificity what steps analysts take to connect data from different populations)). Hemphill’s testimony described the process of linking data from different populations as, essentially, writing and running a new, complex query, (Dkt. No. 80, at 56-57, 60-61), but as ICE concedes, writing new, complex queries, or even the creation of new computer programming, may be required under FOIA if all the query or

program does is “produce[] specific requested fields or records contained within a well-defined database structure,” 6 C.F.R. § 5.4(i)(2)(ii), or extract, compile and sort pre-existing data. (Dkt. No. 15-3, at 10-11). That said, the de Castro Declaration, together with Hemphill’s testimony, does suggest that, given the lack of common identifiers for individuals across populations, the query necessary to respond to Plaintiffs’ requests may need to go beyond these routine functions and conduct some type of substantive calculation or analysis to match the identifiers in different populations that correspond to the same individual, thus merging data from these disparate populations to create a comprehensive alien record that did not previously exist. To analogize to a paper records

system, if ICE kept records of detainers and records of removals in separate drawers with separate subfolders, but both sets of records were clearly marked with individual aliens’ names or identifying numbers such that a particular alien’s detainer and removal records could be readily matched without any further research or analysis, merely putting those records together and producing them to Plaintiffs likely would not constitute “record creation” under FOIA. See, e.g., Disabled Officer’s Ass’n v. Rumsfeld, 428 F. Supp. 454, 456 (D.D.C. 1977) (“The fact that defendants may have to search numerous records to comply with the request and that the net result of complying with the request will be a document the agency did not previously possess is not unusual in FOIA cases nor does this preclude the applicability of the Act.”). However, if the paper detainer and removal records used different identifiers for the same alien—as is apparently the case here with respect to ICE’s electronic records—matching a particular alien’s detainer and removal records could not be done without additional research or analysis to reconcile the differing identifiers. Because neither Hemphill nor de Castro has described in detailed, non-conclusory terms

what these additional analytical steps actually require, it is not clear to the Court whether those steps would, in fact, go beyond the data extraction, sorting and compilation process that is “business as usual” under FOIA. 6 C.F.R. § 5.4(i)(2)(ii). Of course, an agency’s affidavit need only “contain reasonable specificity of detail,” Grand Cent. P’Ship., 166 F.3d at 478 (quoting Gallant v. NLRB, 26 F.3d 168, 171 (D.C. Cir. 1994)), and need not describe every technical step involved in its database functions. And giving credence to ICE’s unrebutted facts regarding the lack of common identifiers for a single individual across populations—as the Court must—it does appear that linking person-centric data from different populations requires additional analytical steps of some kind, and thus that the query ICE would need to conduct to create these

links may cross the “all-important line” between searching a database and creating a record. Nat’l Sec. Counselors I, 898 F. Supp. 2d at 270-71 14 However, the record currently before the Court gives the Court pause as to whether it may properly reach this conclusion. While the Court has drawn the foregoing inferences through its parsing of the record, the generalized and vague explanations ICE has provided as to why

14 The Court has considered the declaration of Dr. Clark, which suggests that ICE could use the various identifiers that exist in ICE’s database as “search terms in its queries [of the IIDS] to retrieve, for example, the records for each time it has removed from the country the people it has targeted with detainers,” without requiring that “only a single identifier be used or that every record contain this single identifier.” (Dkt. No. 84-1, ¶ 9(g)). However, even assuming that this were the case, Dr. Clark’s declaration primarily speaks to ICE’s capacity to do this—which is not in dispute—and not the more fundamental question of whether running a query in the manner he describes would constitute the creation of a new record within the meaning of the FOIA case law. producing Plaintiffs’ requested data requires “record creation,” as discussed above, leave the Court with doubt as to whether ICE’s explanations are sufficiently “detailed” and “non- conclusory” to meet its burden of justifying its decision not to produce this data. Nat’l Sec. Counselors I, 898 F. Supp. 2d at 250. Furthermore, the fact that ICE routinely provided the “linked” data Plaintiffs seek prior to July 2016, as well as unexplained inconsistencies in ICE’s

declarations regarding the time and resources required to fulfill Plaintiffs’ requests, see Section IV.C. infra, raise additional questions about whether ICE was legally justified in concluding that its former process for responding to Plaintiffs’ FOIA requests—which it followed as a matter of course for years prior to July 2016, apparently without ever objecting to Plaintiffs’ requests on the grounds that they required unnecessary “record creation”—actually had never been required of ICE in the first place. These questions are difficult to answer based solely on the high-level and conclusory explanations ICE has provided thus far, and without more concrete, particularized facts about what the “queries” necessary to link data from different populations actually do. See Long v. ICE, No. 17-cv-01097, 2018 WL 4680278, at *5, 2018 U.S. Dist.

LEXIS 167222, at *14-16 (D.D.C. Sept. 28, 2018) (“In a case such as this one, where ICE previously has provided fields and data elements in response to virtually identical requests, individualized explanations as to why FOIA does not obligate ICE to produce the same fields and data elements are essential . . . [This] would shed light on the nature of the ‘additional analysis’ required to respond to each request—something the court cannot discern from . . . the conclusory statements [in ICE’s declaration].”). The Court also has concerns about the sufficiency of the record before it given ICE’s evolving explanations for its refusal to produce the data Plaintiffs request—as Plaintiffs correctly point out, ICE’s explanations have evolved over the course of four declarations by three different ICE personnel, and it was not until the evidentiary hearing that it became apparent that ICE searched populations that “exist[] in a grouping outside of the [IIDS],” rather than querying the IIDS as a whole. (Dkt. No. 91, at 2 (citing hearing testimony)). ICE’s failure thus far to provide a single clear, consistent, detailed and non-conclusory explanation as to why producing the data Plaintiffs seek requires record creation leaves the Court with further concerns as to whether the

evidence proffered thus far is sufficient to meet ICE’s burden on this issue. In any event, notwithstanding the concerns expressed above, because the Court has found that, on the current record, ICE has not met its burden of proving its search was adequate, the Court need not decide at this stage whether (assuming a search limited to specific populations within the IIDS was adequate under FOIA) the process of linking disparate data populations in the IIDS constitutes the “creation of a new record” within the meaning of FOIA. C. Undue Burden In her declaration, de Castro asserts that, even assuming Plaintiffs’ requests do not require the “creation” of new records, the time and resources required to compile requested data “meet[s] the threshold of ‘unduly burdensome.’” (Dkt. No. 58-1, ¶ 22). The evidence ICE puts

forth in support of this claim does not establish that it is entitled to summary judgment. FOIA requires that “an agency shall provide the record in any form or format requested . . . if the record is readily reproducible by the agency in that form or format.” 5 U.S.C. § 522(a)(3)(B) (emphasis added). In determining whether the records at issue are “readily reproducible,” a court “may consider the burden on the defendant” agency, Scudder v. CIA, 25 F. Supp. 3d 19, 38 (D.D.C. 2014), proof of which “must be not only compelling, but also demonstrate that compliance with a request would impose a significant burden or interference with the agency’s operation.” Long v. ICE, 149 F. Supp. 3d 39, 55 (D.D.C. 2015) (quoting Public.Resource.Org v. IRS, 78 F. Supp. 3d 1262, 1266 (N.D. Cal. 2015)). “[S]ubstantial weight” must be accorded to agency affidavits concerning the reproducibility and “the ‘technical feasibility’ of providing records in the requested format.” Everytown for Gun Safety Support Fund, 403 F. Supp. 3d at 348 (quoting 5 U.S.C. § 552(a)(4)(B)). “Courts often look for a detailed explanation by the agency regarding the time and expense of a proposed search in order to assess its reasonableness.” Wolf v. CIA, 569 F. Supp. 2d

1, 9 (D.D.C. 2008) (finding that a frame-by-frame review of microfilm requiring an estimated 3,675 hours at a cost of $147,000 to be unreasonably burdensome); see also Hainey v. Dep’t of Interior, 925 F. Supp. 2d 34, 45 (D.D.C. 2013) (holding that it would be unreasonably burdensome to require the agency to search and review every email sent or received by 25 different employees throughout a two-year time period). By contrast, “where an agency vaguely characterizes a search as ‘costly’ and notes that such a search would ‘take many hours to complete,’” courts have “rejected the agency’s claim as inadequate in explaining why the search would be unreasonably burdensome.” Pinson v. U.S. Dep’t of Justice, 80 F. Supp. 3d 211, 217 (D.D.C. 2015) (finding that agency could not demonstrate undue burden without providing, “at

the very least,” the number of files to be searched and the amount of time required to locate documents). ICE did not initially argue, in its motion for summary judgment, that responding to the requests here would be unduly burdensome. (Dkt. No. 15-3). Marla Jones described the time to respond to the Plaintiffs’ FOIA requests as follows: the average response time for one of Plaintiffs’ FOIA requests in FY2016-17 is five to ten employee hours. If the agency were to respond to as many of the data points requested by Plaintiffs that were not agency records, by creating records, making assumptions and conducting calculations or data analysis, or by answering questions, the response time on one of Plaintiffs’ FOIA requests would be as high as 85 hours, and this number would only increase if the scope of the requests continue to expand in size as well as the length of time covered. (Dkt. No. 23-1, ¶ 22). Plaintiffs assert that the 85-hour estimate is refuted by the fact that “for all of FY 2016 and the first quarter of FY 2017,” ICE produced the data requested by Plaintiffs, when Jones estimates it took employees five to ten employee hours to respond. (Dkt. No. 25-1, ¶ 22(h)).15 ICE has not responded to this argument. In her declaration in support of ICE’s renewed motion for summary judgment, de Castro

states that while ICE’s “database is exceedingly large” and “has computational power to conduct all operationally-required data reporting,” its “computer has many times ‘crashed’” while creating “new, multifactorial data connections for Plaintiffs” in the past. (Dkt. No. 58-1, ¶ 23). De Castro indicates that performing such tasks therefore “substantially interfere[s] with the agency’s computer system and therefore has a direct, negative impact on the ICE mission.” (Id.). However, her declaration is vague as to the nature and frequency of the “crashes” and “interfere[nce]” she describes. De Castro also states that it would likely cost “several million dollars” to “develop or purchase an entirely new computer program or system” adequate to handle requests as complex as Plaintiffs’ requests. (Id. ¶ 34). She asserts that “hundreds of hours

are required” to create a “temporary new record” necessary to respond to Plaintiffs’ requests. (Id. ¶ 32). 16 Plaintiffs argue in response that ICE has not explained the discrepancies in the declarations and offered “no explanation” for its estimate that fulfilling Plaintiffs’ requests would take “hundreds of hours.” (Dkt. No. 59, at 4; Dkt. No. 91, at 9 n. 4). In his declaration, Dr. Clark questions how queries that are, as Hemphill testified, outsourced to a database developer who

15 The FOIA requests at issue here were submitted in November 2016, (Dkt. No. 19-11, ¶¶ 39, 54), and responded to in January and February 2017. (Id. ¶¶ 50, 65). 16 ICE also argues that the total number of Plaintiffs’ FOIA requests—reaching 553 and occupying 57% of ICE’s FOIA analysts’ time in 2018—constitutes an undue burden on the agency. (Dkt. No. 58-1, ¶ 21). As Plaintiffs correctly point out, however, their other FOIA requests are not at issue in this litigation. (Dkt. No. 59, at 4). “runs [them] on a separate server,” (Dkt. No. 80 at 60), could crash the operational database management system. (Dkt. No.84-1, ¶ 9(k)). ICE has not responded to these issues. The Court recognizes that it must accord substantial weight to an agency’s declarations as to the technical feasibility and reproducibility of a records request, and that a “plaintiff must offer more than generalities about technical capabilities of generic systems to overcome or raise

questions about that declaration.” Long, 149 F. Supp. 3d at 58. Plaintiffs have not offered any evidence rebutting De Castro’s claims that there are “hard limitations to the ICE database” that cause it to “crash” or that the technological demands of Plaintiffs’ requests cause ICE’s “computation power [to be] exceeded and “substantially interfere with [ICE’s] computer system.” (Dkt. No. 58-1, ¶ 23). However, in light of the unexplained discrepancies in ICE’s declarations, and the genuine issues raised by Plaintiffs that ICE has failed to respond to, the Court does not find, on this record, that ICE is entitled to summary judgment on its claim of undue burden. D. Renewal While ICE’s motion for summary judgment is denied, the Court will allow ICE to submit

a renewed motion for summary judgment that addresses the outstanding factual questions identified in this opinion with respect to the adequacy of ICE’s search, and its “undue burden” and creating-a-record contentions. As to ICE’s argument regarding the creation of a record, the Court notes that, when asked to explain the specific steps required to “link” data from disparate populations so that the Court could determine whether those steps constitute the creation of a new record, counsel to ICE stated that he was unable to do so, and raised (for the first time in these proceedings) that ICE considers those steps “law enforcement sensitive” and, thus, not subject to public disclosure. However, ICE’s counsel stated that ICE would be willing to provide the Court with a supplemental declaration from an ICE analyst providing the requested explanation, so long as the declaration is filed under seal and reviewed in camera. The Court notes that, in this Circuit, the “presumption of immediate public access” to judicial documents, including any “documents submitted to a court in support of . . . a motion for summary judgment,” may “be overcome only by specific, on-the-record findings that higher values necessitate a narrowly tailored sealing.” Lugosch v. Pyramid Co. of Onondaga, 435 F.3d 110, 126 (2d Cir. 2006). Therefore, to the extent ICE seeks to submit all or portions of any supplemental submission under seal, ICE bears the burden of establishing that the law enforcement interest in keeping any portions it seeks to seal out of the public record is sufficiently compelling to outweigh the presumption of public access, and complying with Local Rule 83.13. V. CONCLUSION For these reasons, it is hereby ORDERED that ICE’s renewed motion for summary judgment (Dkt. No. 58) is DENIED; and it is further ORDERED that ICE may submit a second renewed motion for summary judgment by November 9, 2020. IT IS SO ORDERED. Dated: October 9, 2020 . Syracuse, New York Aer ol ca kc nannies BrendaK.Sannes U.S. District Judge