Assessment of the Epstein Files Transparency Act II

Edition label: 2026-10

H.R. 9694 (119th Cong., 2d Sess.) and companion S. 5005 — drafting assessment and recommended modifications

To The sponsors and cosponsors of H.R. 9694 and S. 5005, and their staff
Date September 20, 2026
Subject What the bill gets right, where it fails, and fourteen specific modifications
Texts assessed H.R. 9694 IH and S. 5005 IS, as introduced July 15, 2026, amending the Epstein Files Transparency Act (EFTA), Pub. L. No. 119-38, 139 Stat. 656 (2025). Bill provisions are cited bare (§ 4(d)); provisions of the enacted Act as EFTA § 2(c) or by context.

BLUF — Bottom Line Up Front

EFTA II supplies an enforcement mechanism: the Act contains no cause of action, penalty, or deadline consequence of its own,1 leaving compliance to the Department absent outside judicial enforcement. The bill’s three most important provisions — the privilege bar (§ 2(b)(2)), the per-redaction justification (§ 2(d)), and the state-officer access-and-enforcement provisions (§ 4(a), (c)) — are the right architecture (M-3, M-4, and M-9 qualify them). Four changes would materially improve it:

  1. Do not repeal the Act’s only classified-accounting trigger — striking § 2(c)(1)(E) orphans the unclassified-summary duty, while § 2(c)(4)’s Federal Register log misses the classification actions in the production. This is the highest-priority low-cost repair because it preserves enacted accountability without M-14’s expansion and classified-handling system (M-1). Intelligence-community holdings are the separate § 2(a) possession question (M-14).
  2. Nothing self-executes — every consequence requires a plaintiff and a favorable ruling (M-2).
  3. The monetary remedies have no sovereign-immunity waiver (M-3).
  4. The production mandates are not reconciled with the other law governing the same records (M-4).

In this memo. Part I identifies eight strengths. Part II ranks and details fourteen modifications in its opening table. Part III states open questions and prior conclusions; Part IV gives eight technical corrections; Position closes with the assessment and its basis.


Part I — What the bill gets right

1. The common-law privilege bar, § 2(b)(2) — the highest-value line in the bill. None of EFTA § 2(c)(1)’s five withholding bases is a privilege, yet the Department of Justice (DOJ) redacted or withheld roughly 200,000 pages on privilege grounds — its own figure2 — arguing Congress had not clearly abrogated the common law. Section 2(b)(2) forecloses that argument by name; adding “law enforcement privilege” anticipates the next one. (Its trailing subsection-(c) clause needs the Part IV, item 3 fix.)

2. “Each redaction,” § 2(d). Enacted § 2(c)(2)’s “[a]ll redactions must be accompanied by a written justification” admits a reading where one aggregate notice covers an entire production. “Each” removes it.

3. Independent causes of action with fee-shifting — state officers (§ 4(a)(1)), victims (§ 4(a)(1) and, differently drafted, § 4(b)(3) — see Part IV, item 5), Members (§ 4(e)(2)). Fee-shifting makes the rights real for plaintiffs without institutional resources. One gap: only § 4(a)(2) and § 4(b)(3)(B) carry expedite instructions; § 4(e), the most time-sensitive route, has none (M-11).

4. Displacing the Touhy regulations, § 4(a)(4) — 28 C.F.R. §§ 16.21–16.29 would otherwise have swallowed the state-officer access rights (limits at M-4; reconcile with § 4(a)(5) per Part IV, item 1).

5. Barring in-camera-only production, § 4(c)(4) — forecloses the look-but-not-keep containment move. The caveat is ours, not the bill’s: the bar cannot lawfully extend to child sexual abuse material (see M-4).3

6. Retroactivity to pending investigations, § 4(a)(6) — drafted emphatically enough to be hard to litigate around.

7. Redaction as evidence of victim status, § 4(g)(1)(A)(i) — a person cannot ordinarily prove they appear in a file they may not see; the concealment becomes the qualifying evidence.

8. Enforcement by state sovereigns — the structural innovation, and the part we would not have thought of. Our prior analysis produced seven reform proposals, none of which enlists a state sovereign.4 Where the executive controls the information, oversight runs through bargaining in which escalation carries visible political risk for Congress;5 a state attorney general sits outside that bargain — no floor time, no comity with the Department, an electoral base of their own in the 43 States that elect the office,6 and an independent criminal docket that gives the records a use. Two structural expectations of ours, stated so they can fail: no attorney-general filing in the first year after enactment falsifies the supply claim, and the pace from § 4(a) suit to enforceable order measures the leverage claim. The supply of willing challengers should also covary with party alignment — under unified government the likeliest plaintiffs are opposition-party attorneys general — though covariation alone cannot distinguish this account from partisan selection. The victim route (§ 4(b)) is a comparator, not an alignment-independent control: eligibility, standing, resources, and representation differ. Record eligible claims, standing dismissals, access to counsel, filings, and time to order for both routes; no AG filing tests supply, not its cause. This changes leverage, not immunity from influence: a state attorney general must still win and enforce a federal judgment.


# Modification Consequence Difficulty Depends on
M-1 Retarget § 2(c)(3) after the (E) repeal; widen § 2(c)(4); settle § 4(d)’s referent Critical Low —
M-4 Reconcile with Rule 6(e), § 3509(m), CIPA, sealing orders Critical Medium —
M-12 Retrospective conformity: designate and re-justify the existing production Critical Medium M-2, M-5
M-3 Sovereign-immunity waiver + fixed damages High Low —
M-5 Machine-readable text + withholding index High Low, conditional on field availability and reconstruction effort —
M-7 § 3 accounting, certification, GAO validation, takedown log High Medium —
M-2 Deemed waiver on missed justification High Low M-5, M-7
M-13 Appropriations conditioned on the M-7 certification High Medium M-7
M-9 Standing per plaintiff route High Low —
M-8 § 4(d) court-supervised process + clearances Medium Medium M-1
M-10 § 4(f) predicates + non-DOJ consequences Medium Low —
M-11 Smaller items (incl. § 2(c)(1)(C); chambers) Medium Low —
M-6 Analytical capacity outside the executive chain Medium High —
M-14A Standardized intelligence-element certification Medium Medium —
M-14B Conditional extension to intelligence-community holdings High if the material exists High M-1, M-14A

If only three changes are possible: M-1, M-4, M-5. Repair operability, reconcile legal collisions, then build the index M-12 requires, regardless of cost. M-12 is Critical but depends on M-5, which works independently. The table ranks consequence; this selection ranks dependency; the BLUF identifies four improvements to the bill outright.

M-1. The repeal strips the trigger from the Act’s only classified-accounting duties

Problem. EFTA publishes unclassified records (§ 2(a)), permits redaction of segregable portions, including classified passages (§ 2(c)(1)(E)), and requires accounting: § 2(c)(3) compels maximum possible declassification and an unclassified summary for each classified item withheld or redacted; § 2(c)(4) requires Federal Register publication of classification decisions, dates, authorities, and unclassified justifications.

Striking (E) leaves § 4(d)(5) cross-referencing § 2(c)(3) for public release of classified information, with three problems. (1) Nothing expressly retargets § 2(c)(3)’s declassify-or-summarize duty after its trigger disappears.7 (2) § 2(c)(4) reaches decisions “to classify” after July 1, 2025, excluding sustained classifications and § 2(c)(3) determinations. The 2025 review kept three documents classified and declassified three (Bates EFTA01730036); neither action is a decision “to classify.”8 (3) § 4(d)(1) orders access to “classified records described in section 2(a),” although § 2(a) describes unclassified records. If this means classified portions of covered records, say so rather than leave the referent to litigation.

Recommendation.

M-2. Make something self-execute

Problem. Section 4(e)(1)‘s 7- and 30-day deadlines merely enable suit. Enforcement depends on plaintiffs’ time and budgets against the Department’s resources, without a deadline consequence.

Recommendation. Deemed waiver. If a redaction’s § 2(c)(2) justification is not both published in the Federal Register and submitted to Congress within a fixed period, deem its EFTA withholding basis waived. Never waive § 2(c)(1)(A), (B), or (D) protections — victim privacy, child sexual abuse material, death-or-injury imagery — for a reporting default. Reach (C), the Department’s discretionary assessment, and other EFTA-supplied bases, but not external constraints: classification decisions, grand-jury secrecy, court orders, or another statute’s prohibition. Require the Department to state those bases or move the governing court (M-4). EFTA’s declassify-or-summarize duties retain M-1’s backstop. Timely publication with late congressional submission gets a short cure period. Waiver changes legal status without an assessor; a court is needed only against refusal to release material with no remaining basis. M-7’s submission log exposes defaults. A daily civil penalty does not self-execute: it needs an assessor, rate, notice, contest, and collection path; even the Clean Water Act assigns penalty-setting to a court.9 Supply that machinery or drop it.

M-3. The monetary remedies have no sovereign-immunity waiver

Problem. §§ 4(a)(1) and 4(b)(3) authorize “monetary relief” against the Attorney General in an official capacity — a suit against the United States, which requires an unequivocal express waiver of sovereign immunity. The Administrative Procedure Act (APA) waiver excludes money damages, and the Supreme Court requires waivers reaching money to be unambiguous.10 The bill identifies no liable sovereign, waives no immunity, defines no damages measure, names no payment source, and sets no limitations period. The counterargument — that authorizing “monetary relief” is itself the waiver, with Bowen v. Massachusetts reading the “money damages” exclusion narrowly — is precisely the litigation an express waiver forecloses.

Recommendation. Keep equitable and declaratory relief under the APA waiver. For official-capacity money relief, draft an express waiver and fixed per-violation damages, deleting “monetary relief” only if sponsors decline to supply them. Draft the individual official’s personal-capacity penalty separately (M-10). Add “disgorgement” only if the bill identifies recoverable gains.

M-4. Reconcile the access provisions with the law governing the same records

Problem. § 4(c) requires production of, and copies of, material withheld under every § 2(c)(1) basis; § 4(e)(1)(A) grants unredacted access “notwithstanding … any other provision of law.” Neither is reconciled with:

Recommendation. State, per regime, whether EFTA II displaces it, is subject to it, or needs a court order reconciling the two. Codify the grand-jury path per the bullet above. Carve § 2(c)(1)(B) material into § 3509(m) inspection, and extend § 4(c)(3)’s carve-back beyond victim personal information to the (B) and (D) material. Make a protective order a precondition of § 4(c) production, and make § 4(b)(2)(A)’s victim-identity redaction mandatory (“shall”), not permissive. Supply eligibility, clearance, storage, and handling rules for § 4(d), borrowing CIPA’s named procedures — incorporation by reference alone supplies no eligibility or clearance apparatus. Narrow § 4(e)’s “notwithstanding” clause so it does not silently override grand-jury secrecy and victim-identity protection.

M-5. Require a usable production format and a withholding log

Problem. EFTA § 2(a) requires “a searchable and downloadable format.” Of roughly 1.38 million cataloged documents, 957,490 carry a native text layer and 423,206 do not (420,903 of them PDFs)14 — roughly 30% silently invisible to text search.

Recommendation. Require machine-readable text throughout — agency optical character recognition (OCR) where needed, validated against a stated accuracy floor — and a structured document-level index linking date, custodian, Bates range, stable identifier, and § 2(c)(1) basis per redaction. The index enables audit, M-12’s cure, and M-2’s waiver. Existing review infrastructure may reduce implementation cost; the Low difficulty rating assumes consistent exportable fields and limited reconstruction effort. Assess field availability and estimate reconstruction effort before treating those assumptions as established: the January 30 letter confirms a document management system, not those fields or costs.15 Cover audio, video, and image redactions expressly in the index and § 2(c)(2): blurs, bleeps, and crops need legal bases. The January 30 letter reports more than 2,000 videos and 180,000 images and describes a victim-protection rationale for redacting videos and images. But the reporting documents examined — the February 14 Section 3 Report and the December 19 and January 30 compliance letters — provide no item-level media-redaction ledger tying each edit to its asserted legal basis.16

M-6. Fund the capacity to read what the bill gives access to

Problem. § 4(e)(1)(A) gives a Member unredacted access to roughly 1.38 million documents within 7 days (the corpus measured at M-5), and § 4(c)–(d) give state officers comparable access; neither the Member nor the officer has any funded capacity to process it — reproducing the executive’s advantage one level down.

Recommendation. Authorize and fund an analytical capability outside the executive review chain — housed in the legislative branch, or contracted under transparency-of-method rules (provenance to the source record, reproducible retrieval, human adjudication of adverse claims) — able to ingest a production and deliver structured cross-document analysis within statutory deadlines. Disclosure: this restates a proposal we published before the bill existed17 (see Part III, prior conclusions).

M-7. Reach the § 3 compliance report

Problem. The bill leaves § 3 untouched: the Department writes and validates its own compliance report, decides when the release is “complete” (which is § 3’s own trigger), and owes only a category-level listing of withheld records (§ 3(1)) — no Bates ranges, no per-document withholding basis.

Two measurements from our corpus audit illustrate the gap. Among 2,670 documents selected by redaction-phrase queries, 569 — 21.31%, a share fixed by the query design, not a prevalence — carry visible legacy-framework markers; a visible marker does not prove the Department’s operative basis.18 And the Department’s own figures leave roughly 2.5 million pages unaccounted for between its more-than-6-million-page potentially-responsive pool and its released production — a gap in an over-collected pool, not established withheld responsive material.19 Require public accounting to reconcile both gaps.

A bipartisan Senate request models independent validation. On March 11, 2026, Senators Merkley, Murkowski, Luján, and Durbin asked the Government Accountability Office (GAO), a legislative auditor, twelve questions about DOJ redaction protocols: staffing, divergent victim/“possible abusers and co-conspirators” directives, software, contractors, and approval chains.20 Citing contemporaneous reporting, the letter asserts DOJ removed thousands of published records with “no apparent explanation.”21 Silent withdrawals destabilize even accurate § 3 reports.

Recommendation. Amend § 3 to require:

M-8. § 4(d) classified access — reframe procedurally, then supply the apparatus

Problem. Section 4(d)(3)’s bar on objections “on the grounds that the material is classified” invites an Article II challenge. M-1 must first settle its referent. Egan’s holding is statutory: its judicial-deference rule applies “unless Congress specifically has provided otherwise.”23 Its statement that classification power “exists quite apart from any explicit congressional grant” is answered partly by M-14’s JFK precedent, but that Act retained a presidential withholding valve; a provision leaving none carries greater exposure. Bush’s signing statement also objected to congressional access,24 supporting the proposed security apparatus for § 4(e) Member access. Reframing cannot supply a clearance: courts cannot grant one (E.O. 13526 § 4.1, cited at M-4).

Recommendation. Make § 4(d) a court-supervised access process: classification governs handling, not refusal. Move § 4(d)(5)’s construction rule first; supply eligibility, clearances, storage, and filing rules using CIPA’s named procedures. Route § 4(e)(1)(A)’s seven-day unredacted Member access through committee secure facilities under chamber rules: its “notwithstanding section 2(c) or any other provision of law” language reaches every § 2(a) record, more plainly still with M-14. Otherwise exclude classified records from that route and use § 4(d). Cite Egan’s statutory frame in the section-by-section.

M-9. Standing — draft per plaintiff route

Problem. A State suing “on behalf of … the residents thereof” (§ 4(a)(1)) acts as parens patriae, protector of residents rather than an injured party. Massachusetts v. Mellon bars that route against the United States, as restated in 2023;25 its application here remains open.26 Own-injury State suits are distinct and stronger. Define whether district attorneys represent “the State.” Victims are the strongest route. Institutional authorization and injury must match: McGahn’s committee enforced a full-House-authorized subpoena; Bethune-Hill’s chamber lacked authority to represent the State and injury of its own.27 Individual Members face Raines v. Byrd: statutory permission to sue did not suffice.28 Congress can make concrete harms actionable, but injury in law is not injury in fact.29 TransUnion’s public-disclosure distinction favors the bill’s proximity to FEC v. Akins. Several § 4(a)(3) harms supply the required downstream consequences; purely informational items still need concreteness.30

Phang v. Blanche supplies preliminary, plaintiff-specific APA relief, with finality conceded and appeals pending. Content questions remain unresolved; an order is not completed public production. This relief does not supply all proposed EFTA remedies: express plaintiff routes and fees, M-2’s litigation-reducing defaults, and M-3’s damages waiver.31

Recommendation. Separate own-injury State suits from resident-representative suits, tying relief to an active State investigation and the State’s own injury. Authorize a suit by a committee authorized by its chamber to vindicate a specified institutional information or subpoena injury. Neither committee designation nor concurrent authorization substitutes for institutional injury; a standing holding supplies neither a cause of action nor an answer on immunity. Keep § 4(a)(3), enumerating the concrete-consequence harms as the pleading hook, citing TransUnion and Akins.

M-10. § 4(f) criminal penalties — restructure or remove

Problem. § 4(f) attaches liability under 18 U.S.C. §§ 1001, 1505, 1519, and 2071 to officers of the Department, the FBI, and the U.S. Attorneys’ offices, naming the Attorney General and FBI Director — statutes only the Department of Justice can charge. This risks under-enforcement, not legal unenforceability. The four offenses have different act, custody, proceeding, and materiality elements; “as applicable” does not map them to conduct. The bill preserves their intent elements, and § 4(f)(2)–(3) closes only part of §§ 1505 and 1519.

In Iran-Contra, accountability failed on durability, not initiation: prosecutions for withholding from Congress produced pleas and convictions, but pardon or reversal undid all except Secord’s result.32 Reversal, classification procedure, charging discretion, and pardon each limit durable accountability.

Recommendation. Map each prohibited act to a specific predicate with its retained elements. Then add two consequences that need no charging decision. First, a civil penalty against the individual official, drafted as its own personal-capacity claim with service, defenses, and collection machinery; the bill’s existing actions are official-capacity and cannot carry it (see M-3). Second, mandatory referral-and-report: a § 4(e)(3)(C) finding that a justification was materially false is transmitted to the Judiciary Committees and published, and the Department must report whether it opened an investigation and, if not, why. Congress cannot compel prosecution; it can make declination visible. The private-prosecutor route is not a template: Young v. United States ex rel. Vuitton concerns contempt of a court’s own order, not statutory offenses.33

M-11. Smaller items

M-12. Cure the existing production — retrospective conformity

Problem. § 2(d)’s per-redaction duty is prospective, and the released production does not comply with it: a pre-enactment FBI instruction directed reviewers to apply bare boxes (“Do not add any exemptions; just redaction boxes”), and material from four inherited record or redaction groupings kept its original redactions, under no § 2(c) re-check the record discloses.37 The consequence: no outside party can tell an authorized victim-privacy redaction from one shielding a person the statute forbids protecting. (Enacted § 2(c)(2) already required justifications at release, so this cures a documented default rather than imposing a new retroactive burden; an express retroactivity clause raises no serious Landgraf problem.38)

Recommendation. Require retrospective conformity: the M-5 index and signed completion certificate within 90 days of enactment, then a 180-day conformity period from certified completion to designate each (A), (B), or (D) redaction and re-justify every redaction under § 2(c)(2). Publish through the index with a Federal Register availability notice; enforce the index deadline separately. For undesignated-and-unjustified redactions, waive only waivable EFTA bases at conformity expiry. For designated-but-unjustified redactions, by that same outside date the Attorney General must file for expedited in camera review in an available authorized proceeding under M-7 on M-11’s affidavit-gate model; otherwise issue M-7’s signed protected-withholding determination, preserve the material, and publish its non-identifying determination or summary in the M-7 log, cross-linked in the M-5 index. This administrative deadline is independent of M-7’s court clock: 30 calendar days after proper filing in a court with jurisdiction. Waive only waivable bases at the applicable determination, judicial disposition, or missed administrative or judicial deadline. The Attorney General must release the lawful remainder within 30 days of the first applicable event; no judicial disposition is required where no proceeding exists. Preserve M-2’s savings and M-4’s reconciliation: no default waives (A), (B), (D), classification, external law or governing orders. Before re-release, apply M-7’s screen; release only the lawfully releasable version. Log results in the M-5 index and M-7 report. Missing designation, justification, determination, publication, screening or review deadlines remains a published, § 4-enforceable default subject to plaintiff eligibility and Article III. For court-ordered redactions, require the Attorney General to move the issuing court within the conformity period on M-4’s December 2025 template. M-7’s basis distribution exposes boilerplate justifications.

M-13. Condition appropriations on certified compliance

Problem. The bill depends on plaintiffs and courts; § 4(f) also needs DOJ charging. The Constitutional Imbalance research paper treats criminal contempt, civil litigation, and appropriations as routes with different dependencies, and records private judicial compulsion.39 Contempt depends on an executive prosecutor; Congress controls appropriations, which this bill leaves untouched.

Recommendation. Make a defined percentage of a named Department account — administrative and leadership-office funds, never investigations, prosecutions, victim services, or § 2 review — unavailable for future obligations when M-7’s certification condition fails. Apply M-7’s first-quarter transition and fifteenth-day window: the preceding valid certificate survives through the deadline; test the new one the next day. Nonfiling, qualification, or withdrawal suspends future obligations; qualification and withdrawal operate immediately, including mid-quarter. A compliant filing or correction restores availability prospectively, never curing obligations incurred during unavailability. Knowingly false certification incurs M-7’s perjury exposure and M-10’s referral-and-report duty. Require the Comptroller General to transmit an adverse GAO report to both Judiciary Committees and both Appropriations Committees within seven days of issuance, recommending expedited consideration. Transmission compels neither committee action nor a floor vote. Unavailability stays tied to certification; GAO informs, since legislative control of execution violates Bowsher v. Synar.40 Unavailable funds cannot lawfully be obligated without a plaintiff or judgment; the Antideficiency Act, 31 U.S.C. § 1341, is the sanction backdrop. Define restricted offices by office and function, never individuals; name the account and add severability (M-11). Removing the standing limitation requires express language in a later appropriations act, making reversal visible and attributable.

M-14. Extending the Act to intelligence-community holdings

Problem. EFTA § 2(a) requires unclassified records in DOJ possession. The possession limit excludes otherwise responsive intelligence-community records not in DOJ possession, even if “unclassified” is struck; otherwise covered records already in DOJ possession remain within the Act.41 The National Security Agency (NSA) neither-confirm-nor-deny response on intelligence material (EFTA00606101) and Central Intelligence Agency (CIA) response (EFTA01089565) establish neither responsive classified holdings nor a service relationship. CIA reported no located records in its acknowledged-affiliation search (November 5, 1999–July 25, 2011), separately neither confirming nor denying a classified connection.42 Holdings beyond DOJ are our disclosed assessment, not established fact; M-14A tests for them before M-14B expands scope.

Recommendation. Split M-14 into independently enacted stages.

The President John F. Kennedy Assassination Records Collection Act of 1992 is precedent, not a clean analogy: it let the President certify postponement instead of defying the statute.47 An extension without that valve invites the confrontation the Act avoided; retain a protective valve and independent congressional scrutiny of the internal executive validator, subject to M-8’s residual exposure.


Part III — Open questions and prior conclusions

Open questions. None must be resolved before drafting:

Prior conclusions, disclosed. M-6 restates proposal 6 of our earlier work;50 other recommendations do not depend on it. Ranking intelligence-adjacent material outside DOJ possession as the most consequential excluded category is our inference from the companion research note’s evidence review, not a ranking either work makes. M-14A tests whether covered holdings can be located; their significance and the intelligence-adjacent assessment require examining the records. Stage A is independently useful; Congress separately decides M-14B. No other recommendation depends on the inference.


Part IV — Technical corrections

  1. § 4(a)(4) vs § 4(a)(5). “Exclusive procedural framework” conflicts with “cumulative” remedies. State which controls, for which plaintiffs and which routes.
  2. § 4(f)(1). The bill specifies fines or imprisonment under 18 U.S.C. §§ 1001, 1505, 1519, and 2071, as applicable, retaining their required mental states. Clarify how each listed act maps onto those offenses and their remaining elements, as M-10 recommends.51
  3. § 2(b)(2). “Except as expressly permitted under subsection (c)” permits nothing — subsection (c) creates no privilege basis. Strike the clause, or say what it permits.
  4. EFTA § 2(a). “Subject to subsection (b)” should also name subsection (c), where the withholding rules live.
  5. § 4(a)(1) / § 4(b)(3). Victims hold two differently drafted causes of action, with different scope and predicate acts. Relief is materially identical; venue is identical in the Senate text but differs inside the House’s — § 4(a)(1) drops the “United States” that § 4(b)(3)(A) carries (the divergence M-11 reconciles). Merge the routes, or state how they interact.
  6. Undefined terms. Section 4(g) defines the bill’s named terms for § 4 only. Define Act-wide “completion,” “other authorized State officer,” “unlawful withholding,” “materially false,” “temporary,” and the reach of “that relate to”; item 8 separately states the substantive scope instruction for that last phrase.
  7. Affirmative segregability, EFTA § 2(c)(1). The enacted paragraph permits withholding or redaction of segregable portions under (A)–(E). Add, after the subparagraph list and before paragraph (2): “The Attorney General shall release every reasonably segregable portion not authorized to be withheld under this Act.” Conform this added duty to M-4’s reconciled protections alongside § 2(a). Capitalize “General” in paragraph (1)’s opening sentence.52
  8. “That relate to,” EFTA § 2(a). The Act’s widest scope phrase, and undefined. The fix runs toward disclosure, not away from it: define what connection to the investigations suffices — in an Act-wide definitions provision, since § 4(g) is scoped to § 4 — rather than leaving the Department to read the phrase for itself.

Position

Our document-level research ranks intelligence-adjacent material outside DOJ possession as the most consequential excluded category; otherwise covered intelligence records in DOJ possession remain within the Act. Classification is a separate withholding question; DOJ reports never invoking that basis.53 The NSA and CIA letters, EFTA00606101 and EFTA01089565, neither establish responsive classified holdings nor a service relationship; CIA’s negative acknowledged-affiliation search is distinct from its classified-connection non-confirmation.54 The Constitutional Imbalance research paper rests its EFTA analysis on executive custody of records, not on any intelligence operation in the underlying matter. Its separate research note on the intelligence-network evidence documents arranged contacts, a private financing structure, and a former NSA officer’s attributed public assessment, not declassified official identification; it assigns no aggregate probability and finds that these records do not establish a service-directed intelligence operation.55 The reviewed record contains neither formal state-service documentation nor independent foreign-government acknowledgment. The first is disclosure-reachable if held in the United States; records outside DOJ possession need M-14’s expansion, while classification separately requires protected handling. The second remains beyond any United States disclosure statute. M-14A tests whether covered holdings can be located; their significance and the underlying assessment require examining them. The diagnostic is independently useful; Congress separately decides M-14B. The drafting rule applied throughout Part II: resolve every ambiguity toward forcing covered material into the open, or forcing a named official to refuse it on the record — under a stated legal basis, on a stated date, subject to independent congressional scrutiny, with M-14A’s operative findings assigned to an internal executive validator.



Bibliography

Cases, statutes, bills, and archival production documents are cited in full in the notes, per legal-citation practice.

Bush, George H. W. “Statement on Signing the President John F. Kennedy Assassination Records Collection Act of 1992.” October 26, 1992. In Public Papers of the Presidents of the United States: George Bush, 1992–1993, Book II, 2004–5. Washington, DC: Office of the Federal Register, 1993. https://www.govinfo.gov/content/pkg/PPP-1992-book2/pdf/PPP-1992-book2-doc-pg2004-2.pdf.

ElSheshai, Alan. National Security Power and Constitutional Imbalance: Executive Powers, Congressional Oversight, and Democratic Trust. Constitutional Imbalance research paper, v8.20.3, 2026.

ElSheshai, Alan. Research Note: Intelligence-Network Underlay. Unpublished research note, September 30, 2026.

Merkley, Jeffrey A., Lisa Murkowski, Ben Ray Luján, and Richard J. Durbin. Letter to Orice Williams Brown, Acting Comptroller General, U.S. Government Accountability Office. March 11, 2026. https://www.lujan.senate.gov/wp-content/uploads/2026/03/GAO-Epstein-Files-Redactions-Review-Letter.pdf.

National Association of Attorneys General. “Attorneys General.” Updated January 5, 2021. https://www.naag.org/attorneys-general/. Archived August 25, 2026, at https://web.archive.org/web/20260825190013/https://www.naag.org/attorneys-general/.

U.S. Department of Justice, Office of the Deputy Attorney General. Compliance letters to the House and Senate Judiciary Committees under the Epstein Files Transparency Act. December 19, 2025, and January 30, 2026.

U.S. Department of Justice. Report under Section 3 of the Epstein Files Transparency Act. February 14, 2026.

Walsh, Lawrence E. Final Report of the Independent Counsel for Iran/Contra Matters. Vol. 1. Washington, D.C., 1993.

<!– Footnote definitions. Pandoc renders these as page-bottom notes; this block produces no heading or body text in the export. –>