The ICE
Accountability Act
A structural analysis of who S. 3891 makes answerable, what it can change, and what it still leaves unresolved.
The bill builds a serious new structure of immigration-enforcement oversight. Its strongest achievement is institutional: it creates a channel of answerability outside the executive branch. Its central limitation appears at the scale of the individual person harmed.
This analysis applies two Toward Reconstitution instruments, the Answerability Index and the Taxonomy of Structural Harm, to the introduced text of S. 3891. It reads the bill on its own terms as a genuine attempt to build oversight where existing arrangements failed, then asks where that architecture still routes harm through procedures that can look fair while leaving the underlying relation intact.
Neither instrument is a legal assessment. Both are ways of making a structural argument visible: where power sits, whose feedback can alter outcomes, who can act, who must answer, and what happens after a system causes harm.
What does the bill actually change?
The ICE Accountability Act creates the Commission for Independent Monitoring of Immigration Enforcement, a four-member body structurally separated from the executive branch. Two monitors are appointed through House leadership and two through Senate leadership, with each appointment requiring consent from the opposing party's leader. Members must bring relevant expertise and may hold no other federal, state, or local office while serving.
The Commission can observe immigration agents during operations, conduct unannounced site visits, and access Department of Homeland Security records without prior notice, including body-camera footage, training materials, and civil-rights complaints. It can issue formal findings, refer matters for criminal prosecution, hold public hearings, and bring federal civil actions seeking penalties for continued noncompliance.
Section 4(c) matters structurally because an individual agent's noncompliance is imputed to the supervising agency. Institutional responsibility cannot simply dissolve into the conduct of a single employee. The bill also creates a public complaint portal, protects whistleblowers, requires rapid notification of specified critical incidents, and physically separates the Commission from executive-branch office space.
What it does not change is the substantive law of immigration enforcement. The underlying authority to arrest, detain, and deport remains intact. The bill changes who watches that power, what information the watcher can obtain, and what consequences can follow a finding of noncompliance.
Who becomes answerable to whom?
Every oversight system answers a prior question before it answers its stated one: accountable to whom, and in which direction does that account flow?
Immigration enforcement has largely answered upward and inward. ICE and CBP sit inside the Department of Homeland Security, which sits inside the executive branch that directs enforcement. The internal offices meant to check that chain remain inside the same institutional structure they are meant to scrutinize. The people arrested, detained, and deported by those agencies do not possess an equivalent reciprocal channel back into it.
S. 3891's structural proposal is to create a second channel in Congress, outside the branch conducting enforcement. Through the complaint portal, people affected by enforcement also gain a route into that channel. The consequential question is therefore larger than whether ICE and CBP should comply with law. It is who gets to determine whether they did, and whether that determination can survive the will of the branch being checked.
What shape does the bill create?
The Answerability Index reads eight functions from dominion logic at the center to relational logic at the rim. The point is not the average. The unevenness is the finding: which functions move outward, which stay concentrated, and what the distance between them allows the system to do.
Eight functions
Select a function to see what the score is anchored to in the bill's text.
The bill learns better than it repairs. That inversion is the most important thing the shape reveals.
The highest and lowest spokes
Adaptation scores highest at 7. The sunset-and-reversal mechanism is a genuine structural rarity: the oversight structure can withdraw once sustained compliance is demonstrated and return if conditions regress.
Agency scores lowest at 2.5. The person the bill is meant to protect can provide information, but has almost no lever of their own once that information enters the system.
A strong function can make a weak one easier to tolerate
Strong Adaptation creates a political argument for accepting weak individual Agency: if the institution can correct itself, individual standing can be treated as less necessary. The high spoke does not cancel the low one. It helps explain how the low one becomes tolerable.
Feedback and Accountability move together
Feedback at 6 and Accountability at 5 are comparatively aligned. Public complaints are statutorily relevant to findings, and findings can trigger real institutional consequences. That is substantially stronger than a system that collects signal and routes it nowhere.
The qualification is scale: the accountability on the other end of that feedback is primarily institution-to-institution, not person-to-institution.
Repair and Adaptation invert the common failure
A common system repairs individual cases while leaving the conditions that keep producing them intact. This bill does almost the reverse. Adaptation is strong at 7; Repair is weak at 3. The structure can learn and change while harm already experienced by a specific person remains largely unrestored.
Who scored it remains part of the finding
This shape is scored from the bill's text by an outside analyst. Someone who has lived through an immigration-enforcement encounter could reasonably score Agency and Repair lower because aggregate protections may not be available at the scale and speed of the encounter itself. The distance between those two scorings would be analytically meaningful.
The scores are arguments anchored to the bill's design, not empirical measurements of how a future Commission would perform. The shape is useful because it makes the argument inspectable.
Where can fair-looking process still route harm?
The Taxonomy of Structural Harm names six mechanisms by which a system can produce harm without requiring anyone inside it to act in bad faith. A well-designed reform can close some mechanisms while leaving others available. That is exactly what happens here.
What can the unit of account not see?
Diagnostic question: What can this system's unit of account not see?
The operative unit is a serious or willful violation found by a three-of-four vote and then reported in aggregate. A single severe encounter that does not fit a larger pattern, or does not clear that evidentiary threshold, can remain structurally faint even when the harm to the person involved is profound.
Where does protection arrive after the fall?
Diagnostic question: Where does improving your position cost you?
Whistleblower protection is meaningful, but reactive. It addresses retaliation after professional risk has already been incurred. The formal protection can therefore exist at the same moment the practical cost of reporting is already being absorbed.
Can the person's situation worsen faster than oversight can move?
Diagnostic question: Does being in trouble generate more of it?
An immigration case proceeds on its own legal timeline. Arrest, detention, and removal do not pause while the Commission reviews information, develops findings, or reaches a three-of-four vote. Institutional review can move more slowly than the consequences it is meant to scrutinize.
Does proving harm depend on the institution accused of causing it?
Diagnostic question: Does entry require prior entry?
A complaint may become actionable through body-camera footage, encounter records, training materials, and other evidence controlled by the agency itself. The Commission has unusually strong access rights, but the person's account can still depend on institutional records existing, being preserved, and being produced in time to matter.
Whose uncounted labor makes oversight possible?
Diagnostic question: What unlogged labor holds this up?
The system depends on complainants documenting harm and whistleblowers accepting professional risk to surface information the Commission needs. That labor is structurally necessary to the oversight process, yet neither the complainant's work nor the whistleblower's risk is counted as part of the cost of making the system function.
The bill refuses to let institutional harm dissolve into one employee
Diagnostic question: Do identical acts get judged by who performed them?
Section 4(c) is a genuine strength. An individual agent's noncompliance is imputed to the agency that supervises them. That resists a common structural escape hatch: isolating failure in one actor so the institution that trained, authorized, supervised, or tolerated the conduct escapes responsibility.
Five of the six mechanisms still have room to operate.
That does not negate the bill's oversight architecture. It shows that oversight and structural harm occupy different layers of the same system. S. 3891 does serious work on one layer while leaving another only partially altered.
Which relationship does the bill actually repair?
Congress ↔ enforcement agencies
The bill breaks the circularity of agencies being overseen only inside the executive branch that directs them. Congress gains an independent structure with access, enforcement capacity, public reporting, and the ability to survive the preferences of a single administration.
Person ↔ enforcement agency
The individual can be heard through the complaint process, but is not given comparable standing inside the remedial structure. The system can answer institutionally without necessarily answering to the specific person who experienced the harm.
Reform changes a structure's procedures while the relation producing the rupture can remain substantially intact. Repair changes the relation itself. Both instruments point to the same seam in S. 3891: the bill repairs the oversight relation between Congress and immigration agencies much more fully than it repairs the enforcement relation between an agency and the person it acts upon.
That distinction is not an argument against the bill. Removing the circularity of executive-branch self-oversight is a genuine precondition for credible accountability, and the sunset-and-reversal mechanism is unusually responsive design. Precision about the distinction simply tells us what kind of accountability has been built, institutional and aggregate, and what kind remains incomplete, individual and restorative.
Where does the structure still stop?
No right to compel individual review
A person affected by enforcement can submit a complaint but cannot require the Commission to investigate that specific case.
No party status
Once the Commission acts, the complainant has no formal role in the proceeding, vote, or continuing participatory right.
No person-calibrated remedy
Penalties and court orders run toward agency compliance. The bill does not create a remedy designed to restore something directly to the individual harmed.
The bill's protective frame extends further than a surface reading might suggest because public complaints and whistleblower disclosures are built into the architecture. But it stops short of giving the person harmed standing inside the remedial process. If the legitimacy of state power is relational, the person governed or enforced against has to appear somewhere in the structure of remedy, not only in the structure of monitoring.
What would stronger answerability require?
As the complaint process is implemented, is there any intention to give an individual complainant standing to request review of their own case specifically, rather than having their account function only as an input to aggregate findings?
Has the Commission's design considered any remedy that runs to the person harmed directly, distinct from the penalty paid by the agency to the government?
Given that substantiating a complaint may depend on body-camera footage and records the agency itself controls, is there a provision ensuring the Commission's access to that evidence cannot be delayed in a way that outlasts an individual's own case?
Scope
This page analyzes the introduced Senate text of S. 3891, introduced February 12, 2026 by Sen. Christopher Coons with Sen. Elizabeth Warren as original cosponsor.
The analysis is structural rather than legal and does not assess the constitutionality or litigation posture of the bill.
Instruments used
The Answerability Index maps how eight functions are organized across a system: Power, Access, Feedback, Accountability, Agency, Care, Repair, and Adaptation.
The Taxonomy of Structural Harm traces six mechanisms through which harm can be produced or sustained by ordinary process: the Ruler, Cliff, Trap, Door, Ledger, and Scale.
Primary material
Analysis by Melissa Cosgrove for Toward Reconstitution. The Answerability Index™ and the Dynamic Fractal Theorem of Relation © 2026 Melissa Cosgrove. This analysis is an interpretive framework for structural evaluation and should not be treated as legal advice or as evidence in any formal proceeding.