Line of Investigation 10: Prosecutorial Scope Evolution
Executive Summary
The SDNY Epstein investigation issued 257 grand jury subpoenas between July 2017 and June 2021, but the distribution was radically uneven: 79 subpoenas (40% of all dated subpoenas) were issued in August 2019 alone — the month Epstein died — while just 4 were issued in all of 2021. The subpoena record reveals an investigation that began with a narrow communications-surveillance predicate (telecom and tech companies, 2017-2019), exploded into a broad financial and institutional dragnet immediately after the August 10, 2019 death, then contracted steadily through 2020 and 2021 until it was issuing fewer subpoenas per quarter than it had issued per day in August 2019. The investigation never subpoenaed 575 entities identified in the corpus, The corpus contains no matched returns from major technology companies (Google, Facebook, Amazon, Apple), and left 124 demand clauses — including 83 classified as HIGH severity — without any identifiable responsive production. By the time of the Maxwell arrest in July 2020, the Epstein-specific investigation had already pivoted from building a conspiracy case to collecting records from airlines, credit agencies, and estate-adjacent entities. The subpoena issuance rate declined steadily from August 2019 through the final subpoena in June 2021.
1. The Six Phases: Subpoena Issuance by Period
The 196 dated subpoenas cluster into six distinct phases, each with a different investigative character.
| Phase | Period | Subpoenas | Clauses | Primary Focus |
| ------- | -------- | ----------- | --------- | --------------- |
| 1. Covert investigation | Jul 2017 -- Jun 2019 | 21 | 100 | Telecom, tech, law firm |
| 2. Arrest month | Jul 2019 | 11 | 74 | MCC, Deutsche Bank, AmEx |
| 3. Death month | Aug 2019 | 79 | 1,080 | Financial dragnet, MCC staff, telecom |
| 4. Post-death 2019 | Sep -- Dec 2019 | 38 | 273 | Financial deep-dive, tech follow-ups |
| 5. Year 2020 | Jan -- Aug 2020 | 43 | 210 | Airlines, Interlochen, continued financial records |
| 6. Year 2021 | Mar -- Jun 2021 | 4 | 7 | Interlochen, attorneys, one airline |
The decline from Phase 3 to Phase 6 is steep and unbroken: 79 subpoenas, then 38, then 43, then 4. Clause counts — a proxy for investigative specificity — follow the same trajectory: 1,080 to 273 to 210 to 7. The investigation's compulsory process output dropped by 99.4% between its peak month and its final year.
2. Pre-Arrest: Building the Communications Predicate (Jul 2017 -- Jun 2019)
The pre-indictment phase was dominated by communications surveillance. Of 21 subpoenas, 15 targeted telecommunications and technology companies.
| ---------- | ------- | ------------- |
| Telecommunications | 10 | AT&T, Verizon, T-Mobile, Century Link, Verizon Wireless |
| Financial Institution | 1 | Gold Coast Federal Credit Union (EFTA00152472) |
| Cryptocurrency / Fintech | 1 | Iterative Capital (unrelated case, see Dossier 08) |
Only one financial institution was subpoenaed before the arrest. The law firm subpoenas — three to Boies Schiller Flexner, Epstein's counsel — signal document collection related to prior legal proceedings, not an offensive expansion of targets. This phase contains the investigation's single most significant temporal gap: 524 days of silence between the initial Google subpoena (July 13, 2017) and the next activity (December 19, 2018), classified as CRITICAL severity in the gap analysis.
3. The August 2019 Explosion and Rapid Contraction
Epstein died on August 10, 2019. In the 21 days that followed, prosecutors issued approximately 70 subpoenas — more than in the preceding two years combined. The August subpoenas ran along three simultaneous tracks:
Track A -- MCC Death Investigation: Subpoenas to correctional staff Tova Noel (
EFTA00075342), Clyde Washington (
EFTA01659608), Perry Joyner (
EFTA01659604), Michael Davis (
EFTA01659606), Samantha DiMisa (
EFTA00092851), S. Shakir (
EFTA00092849), R. Adams (
EFTA00092841), R. Silvia (
EFTA00092843), and Michael Kearins (
EFTA00092961), plus 6 subpoenas to MCC itself. A subpoena to 4chan (
EFTA00123462) on August 13 addressed the leak of death-scene information to that platform.
Track B -- Financial Dragnet: Twenty financial institution subpoenas deploying 12 U.S.C. S 3415 (Right to Financial Privacy Act) in a concentrated burst on August 16-19, targeting Wells Fargo, USAA, TD Bank, Navy Federal Credit Union, MCU, JPMC, Citibank, Bank of America, Santander, and others. This was the only statute cited anywhere in the 257-subpoena corpus.
Track C -- Broad Communications Sweep: AT&T (5 subpoenas), Verizon (3), Sprint, T-Mobile, Comcast, and Apple — capturing phone records, subscriber data, and IP logs.
The death month also includes 4 redacted-target subpoenas, most notably EFTA01688067 (August 15, 2019): 246 pages and 128 demand clauses — the single most detailed subpoena in the entire corpus, with its target completely concealed.
4. The Narrowing: What Changed After August 2019
The post-death phases show a progressive shift from offensive investigation toward records collection from Epstein's own financial ecosystem and institutional connections.
Phase 4 (Sep -- Dec 2019): Financial Deep-Dive
Seventeen of 38 subpoenas targeted financial institutions. New targets included UBS (EFTA00079147), Fifth Third Bank (EFTA00079159), Credit One Bank (EFTA00109004), First Premier Bank (EFTA00109044), and FirstBank Puerto Rico (EFTA00010454). The estate executors — Darren Indyke (EFTA00088984) and Richard Kahn (EFTA00088989) — were subpoenaed on October 18, 2019. Neither produced documented returns. The estate-executor subpoenas reflect a shift toward asset identification — a standard phase in trafficking prosecutions involving forfeiture and victim restitution.
Phase 5 (2020): The Airline Pivot
The single largest cluster of 2020 subpoenas was a January 23-27 airline blitz: 22 subpoenas to seven carriers (United, Southwest, JetBlue, Delta, American, Alaska, Airline Reporting Corp) and Expedia. This was the investigation's most comprehensive attempt to reconstruct travel patterns, but it came 6 months after the death and targeted commercial airlines — not the private aviation companies, charter services, or FBOs through which Epstein primarily traveled.
Five subpoenas went to Interlochen Center for the Arts (EFTA00092436, EFTA00092440, EFTA00020846, EFTA00092633, EFTA00026851) in February 2020 — investigating Epstein's donations to the youth arts school. COVID-19 produced a gap from March through mid-April 2020, and activity never recovered to pre-pandemic levels.
Phase 6 (2021): Effective Cessation
Four subpoenas in 18 months: one more to Interlochen (EFTA00019503), two attorney subpoenas (EFTA00088596, EFTA00080578), and a single final American Airlines subpoena (EFTA01660693) on June 29, 2021. The grand jury's compulsory process had effectively ended.
5. The Returns Lag: When Did Data Actually Arrive?
The subpoena record shows aggressive issuance; the returns record shows delayed and partial fulfillment. The earliest documented returns arrived on August 5, 2020 — a full year after most subpoenas were issued. The largest single production came on September 29, 2020:
| -------- | --------------- | ------ |
| Deutsche Bank | 96,019 (x4 copies) | Sep 29, 2020 |
| JP Morgan Chase | 52,949 (x4 copies) | Sep 29, 2020 |
| American Express | 2,308 (x7 copies) | Sep 29, 2020 |
| AT&T | 2,623 (x16 copies) | Aug 5, 2020 |
| USAA | 489,914 (total across 33 returns) | Various |
The returns from 2021 are dominated by GM-related materials (Ghislaine Maxwell prosecution) — 44,751 pages across 12 returns from March through October 2021. By then, the data production had shifted from the Epstein investigation to the Maxwell trial.
6. What the Investigation Never Asked For
The gap analysis identified 575 entities that appear in the corpus but were never subpoenaed. These include individuals with thousands of document appearances:
| Entity | Category | Document Appearances | Subpoenaed? |
| -------- | ---------- | --------------------- | ------------- |
| Alfredo Rodriguez | associate | 17,187 | No |
| Andrew Stein | politician | 9,723 | No |
| Boris Nikolic | academic | 8,202 | No |
| Brock Pierce | business | 1,446 | No |
| Bill Clinton | politician | 1,458 | No |
| Bill Gates | business | 1,281 | No |
| Bill Richardson | politician | 1,253 | No |
The investigation also never subpoenaed any cryptocurrency exchange despite documented crypto engagement (see Dossier 08), any private aviation company or FBO despite Epstein's extensive private air travel, or any educational institution other than Interlochen (no subpoenas to Harvard, MIT, or other universities that received Epstein funding).
Seven investigative tracks were defined in the prosecutorial query graph. One — Medical & Forensic — received zero subpoenas. The Government / Corrections track went dark after August 20, 2019 (a 40-day active window). Credit Reporting Agencies and Money Transfer Services were each active for only 18 days.
7. Fulfillment: The 45% Black Hole
Overall clause fulfillment across the investigation:
| -------- | --------- | ----------- |
The 45.1% UNKNOWN rate means the investigation either failed to track compliance for nearly half its demands, or compliance records were not preserved in the released documents. Among entities with documented HIGH-severity unfulfilled demands: Google (3 subpoenas, 2017-2019, zero documented returns), Facebook (4 subpoenas, zero returns), Amazon (2 subpoenas, zero returns), Apple (1 subpoena, zero returns), MCC (6 subpoenas, zero returns), FirstBank Puerto Rico (3 subpoenas, zero returns), Wells Fargo (2 subpoenas, zero returns), and TD Bank (1 subpoena, zero returns).
8. The Data Class Shift: From Communications to Banking to Personnel
The types of records demanded changed as the investigation evolved:
| Data Class | Phase 1 | Phase 3 (Death) | Phase 5 (2020) | Phase 6 (2021) |
| ----------- | --------- | ----------------- | ---------------- | ---------------- |
The shift from phone_records (Phase 1) to bank_records (Phases 3-4) to personnel (Phase 5) traces the investigation's arc: from surveillance predicate, to financial mapping, to institutional record-keeping. By 2020, the investigation was collecting employment records and donor histories — the work of asset identification and background documentation, not active conspiracy investigation.
9. Records and Sources Not Represented in the Corpus
The following categories of records are absent from or underrepresented in the EFTA production. Many factors affect subpoena issuance rates over time, including resource allocation, competing cases, COVID-19 disruptions, and staff changes.
Direct subpoenas to associates. Of 575 entities identified in the corpus that were never subpoenaed, many appear in thousands of documents. The estate executors (Indyke and Kahn) were subpoenaed once each, with no matched returns in the corpus.
Private aviation records. The January 2020 airline blitz targeted commercial carriers. Records from charter companies, FBOs, and private aviation services through which Epstein primarily traveled are not represented in the subpoena record.
Offshore financial records. Three subpoenas to FirstBank Puerto Rico produced no matched returns. No subpoenas to banks in the USVI, British Virgin Islands, or other Caribbean jurisdictions where Epstein maintained corporate entities appear in the corpus.
Trafficking and conspiracy statute citations. The only statute appearing in the subpoena corpus is 12 U.S.C. S 3415 (financial privacy). No RICO, no 18 U.S.C. S 1591 (sex trafficking), no money laundering statutes appear in any rider clause. While statutes typically appear on grand jury cover pages rather than riders, their total absence from the documentary record limits the ability to track the evolution of legal theories.
Compliance enforcement records. Seventy-nine subpoenas had no linked returns. Multiple major technology companies received repeated subpoenas with zero documented production. No contempt motions, enforcement actions, or compliance escalations appear in the corpus.
10. Questions for the Record
Were the 79 subpoenas with no documented returns fulfilled through other channels? Were these entities held in contempt? Did they produce to a different case number? Were returns produced through channels not reflected in the EFTA corpus?
What are the 27 redacted subpoena targets? The largest single subpoena in the corpus (EFTA01688067, 128 clauses, 246 pages) has a completely concealed target. Congress can compel disclosure of whether these subpoenas targeted individuals, corporations, or foreign entities.
What accounts for the 524-day gap in subpoena activity? The temporal gap between July 2017 and December 2018 — while the SDNY investigation was nominally active — is the longest period of inactivity in the subpoena record.
Does compliance tracking exist for the 45.1% of demand clauses in UNKNOWN status? The investigation's compliance records may have been maintained but not preserved in the EFTA release, or compliance tracking may not have been systematically documented. Clarification would resolve this ambiguity.
How many named co-conspirators were directly subpoenaed? The investigation identified co-conspirators by name in charging documents and NPA agreements. The PQG shows subpoenas to only two estate-related individuals (Indyke and Kahn) and correctional staff — not to any alleged participant in the trafficking network.
Source Citations
All EFTA numbers link to their hosted documents at justice.gov/epstein/files/. Key structural citations:
- Phase analysis derived from 196 dated subpoenas in
prosecutorial_query_graph.db
- Largest redacted subpoena: EFTA01688067 (128 clauses, 246 pages)
- Final subpoena: EFTA01660693 (American Airlines, Jun 29, 2021)
- Gap analysis: 779 total gaps (575 unsubpoenaed entities, 124 unfulfilled demands, 50 partial responses, 27 redacted targets, 3 temporal gaps)
This dossier is part of the Prosecutorial Query Graph Lines of Investigation series. See 00_INDEX.md for methodology and the complete list of dossiers.