Part One — The Instrument
Three separate mechanisms operate together. The first is a pardon power redefined into class-indemnification: the mass proclamation covering approximately 1,500 individuals connected to January 6, 2021, signed on Trump's first day back in office; the twenty-four pardons of FACE Act defendants issued together three days later; the November 7, 2025 umbrella proclamation for offenses "related to the 2020 presidential election." These are not clemency in the case-by-case sense the pardon power has historically taken. They are class-based indemnifications, and the class is defined by political relationship to the current administration.
The second is the Anti-Weaponization Fund — a $1.776 billion appropriation created by the May 18, 2026 Trump v. IRS settlement, designed to compensate people whose prosecution the administration characterizes as "weaponization." The Fund's own Settlement Agreement cites Biden-era FACE Act enforcement, the 2021 school-boards memo, and IRS targeting of ideological groups as its paradigm examples. The class the Fund would have compensated is, on the FACE Act question, the same twenty-four people whose convictions the January 23 pardons erased. On the January 6 question, it is the class Mark McCloskey's approximately four hundred administrative claims to DOJ, filed in December 2025 seeking one to ten million dollars each, sought to move through the Fund's successor pipe — the Federal Tort Claims Act route, which pays automatically from the Treasury Judgment Fund once DOJ agrees to a settlement, requires no appropriation, and offers no APA rulemaking process for outsiders to challenge.
The settlement's most consequential feature may therefore be procedural rather than financial. Thirty-five former federal judges have now asked the court to reopen Trump v. IRS, arguing that the case was non-adversarial because the President exercised actual control over the government's defense, that the parties' interests were effectively “one and the same,” and that the litigation was pursued in bad faith. Their filing points to evidence that the IRS had prepared defenses to Trump's claims — including a memorandum reportedly recommending dismissal — that DOJ allegedly disregarded while negotiating the settlement. If those allegations are sustained, the Fund was not simply a compensation mechanism produced by ordinary adversarial litigation. The litigation itself may have been part of the mechanism: the government that was supposed to resist the claim became the government negotiating the remedy. Sheldon Whitehouse
The third is the doctrinal architecture. On July 1, 2024, in Trump v. United States, Chief Justice John Roberts's majority opinion established a three-tier framework for presidential immunity: absolute immunity for "core constitutional powers" (the pardon is explicitly named), presumptive immunity for other "official acts," no immunity for "unofficial acts." Justice Barrett's concurrence acknowledged that the accompanying evidence-exclusion rule — barring prosecutors from using evidence of official conduct to prove related unofficial conduct — went further than she would have preferred. Justice Sotomayor's dissent named the practical effect: functionally near-total immunity for anything a president can plausibly characterize as official. Whether a given act is "official" now depends less on its content than on whether the president can claim it was.
The three instruments were built at different times, by different actors, for different immediate purposes. The pardon apparatus was reconfigured in March 2025, when Trump dismissed career pardon attorney Liz Oyer and replaced her with political appointee Ed Martin Jr., who also announced he would re-examine Biden's pardons. The Fund was announced in May 2026 as part of a settlement. The immunity ruling came from a Court whose senior justices have documented undisclosed financial relationships with the political constituency the instrument now protects — a pattern Wartime: Captured establishes in full. None of these three moves required the others to make sense on its own terms. Together they construct an architecture in which one defined class of people faces neither prosecution for past acts nor financial exposure for having been prosecuted for those acts, and the arrangement itself is doctrinally insulated from prosecution.
The arrangement is not less corrupt because it operates through law. It is a different form of corruption. Ordinary corruption typically involves crossing a legal boundary — bribes for votes, contracts for friends, personal enrichment through official channels. This arrangement works from inside the boundary. It uses lawful powers to redefine who is exposed to prosecution, who is exposed to financial liability, and ultimately whose conduct the law is permitted to reach. The distinction matters because the familiar instruments of accountability are designed to detect corruption that violates the rules, not corruption that changes the rules. That helps explain why so much of the reaction to it — Casey Michel's Just Security call for a Church Committee, the three lawsuits against the Fund, the congressional letters — feels off-target. The problem is not simply that individual officials may have abused their authority. The deeper problem is an architecture in which legal authority is being used to construct immunity for a politically defined class, while the mechanisms capable of challenging that construction are themselves placed behind the same legal barriers. This is corruption operating through law rather than outside it.
Part Two — The Class
A protected class is only visible when named. The pardon warrants and the Fund's defined beneficiary categories name it in specific terms. Anti-abortion activists convicted under the FACE Act. January 6 defendants. 2020-election-related offenders. Public officials convicted of corruption in office. Foreign nationals convicted of illegal campaign contributions. And through the extended orbit — friends, business partners, and the loosely-defined circle around the administration itself — a wider set of individuals whose relationship to the family or the movement made their exposure to the state's coercive machinery either avoidable when they needed favors or, on rarer occasions, unavoidable when the machinery did not know to spare them.
The FACE Act class matters as the sharpest single instance because it makes the two-sided instrument entirely legible on one docketed set of people. The Anti-Weaponization Fund's Settlement Agreement, four months after the FACE Act pardons, cited Biden-era FACE Act enforcement as a paradigm example of the harm the Fund was designed to compensate. The class the Fund would have compensated on that question was the same class the pardons had already indemnified — the same names, on the same convictions, from the same offenses. Lauren Handy, sentenced to fifty-seven months. Chester Gallagher, sixteen months. Heather Idoni, twenty-four months across three districts. Twenty-four pardons, one day, one offense category. If the Fund had ever operated, these people would have been its paradigm claimants. The pardon indemnified them going backward; the Fund would have compensated them going forward; both instruments defined their class by the same criterion.
Beyond the docketed classes runs the redemption pattern — pardon followed by formal state role — that turns individual clemency into institutional reintegration. Two instances now sit on the record. Charles Kushner, pardoned in December 2020 (first term) for tax evasion, witness tampering (including a scheme involving hiring a prostitute to entrap and record his brother-in-law), and illegal campaign contributions, was nominated Ambassador to France in 2024 and confirmed in early 2025. Rod Blagojevich, whose fourteen-year federal sentence for wire fraud, conspiracy, and extortion under color of official right was commuted by Trump in February 2020 and then fully pardoned on February 10, 2025, was nominated Ambassador to Serbia later that year. The pattern is: pardon erases criminal conviction from active status; ambassadorial nomination reintegrates the pardoned individual into state authority in a role requiring institutional legitimacy that the pardon alone provides. Two instances is a pattern rather than an anecdote, and both instances involve convictions for unusually egregious conduct that would ordinarily disqualify their subjects from Senate-confirmed foreign service roles indefinitely.
Outside the pardon list runs the extended orbit — the friends and family-adjacent figures whose proximity to the administration becomes usable as either a shield or a lever, depending on which direction the coercive machinery is asked to run. The clearest instance is Paolo Zampolli, the modeling agent who introduced Donald Trump to Melania in the 1990s, now serving as Trump's special envoy and a Kennedy Center trustee. In June 2025, Zampolli called David Venturella, a high-ranking ICE official, about his former partner Amanda Ungaro, a Brazilian model who was in a Miami jail on a workplace-fraud matter and about to be released on bail. According to the New York Times, Venturella then called ICE's Miami headquarters to seize her before her release, emphasizing on the call that the request was "a favor for a friend of the president." Ungaro — who had herself been appointed U.S. Ambassador to Grenada during the first term, and who is the mother of Zampolli's sixteen-year-old son — was later deported to Brazil. Zampolli denies asking for deportation. DHS denies political motivation. Melania's spokesperson said the First Lady "has no knowledge of, nor involvement in, the personal affairs of Mr. Zampolli and Ms. Ungaro."
Even The Protected Class Isn't Untouched
The Ungaro-Zampolli story documents the coercive machinery being made available to those in the extended orbit for personal use. A separate story documents the same machinery operating on the extended orbit when no favor is called in: in November 2025, ICE detained Bruna Ferreira in Massachusetts as she drove to pick up her eleven-year-old son from school. Ferreira had been engaged to Karoline Leavitt's brother Michael; she is the mother of Karoline Leavitt's nephew. Karoline Leavitt is on record publicly defending mass-deportation policy from the White House podium. Ferreira was transferred to a Louisiana detention facility, held for nearly a month, and released on the minimum permitted bond of $1,500 after an immigration judge ordered it. No administration figure intervened. The significance is not that the machinery reached someone in the administration's extended family orbit. It is that it reached her while a member of that family was publicly defending the policy that produced it. Nothing in the record establishes that Ferreira's detention was intended as a warning, or that anyone expected Leavitt to intervene. It may have been ordinary enforcement operating without regard to political proximity. But that ambiguity is itself the problem. A system that can be made selectively available to political allies can also reach the families of those defending it. Whether that is merely structural, selectively enforced, or capable of functioning as coercive signaling is an open question. The machinery does not have to be deliberately aimed at the loyalist to create a condition in which loyalty offers no insulation from it.
Part Three — What the Class Signals
The class defines the political constituency. Reading the pardon list chronologically, held against the Fund's stated beneficiary categories, reveals whose conduct the arrangement was built to protect and what kinds of conduct were considered pardonable. Four categorical signals emerge.
The first is public corruption. Over a dozen pardons in the current term have vacated convictions of public officials for bribery, honest-services fraud, extortion under color of official right, and campaign finance violations. Rod Blagojevich, Michele Fiore, Scott Jenkins, Michael Grimm, John Rowland (twice-convicted), Alexander Sittenfeld, Jeremy Hutchinson, Brian Kelsey, Glen Casada and Cade Cothren, Enrique and Imelda Cuellar (Democratic congressman and his wife — the bipartisan legibility move), Timothy Leiweke, Stephen Buyer, Wanda Vazquez Garced. Historically, pardons for public corruption have been rare, precisely because they signal that the pardoning president does not consider abuse of public office a serious offense. If the pardon power is now being used to erase public-corruption convictions at this frequency, the signal to current officeholders is unavoidable: abuse of office is not, in this administration, considered a punishable offense. That signal itself operates as a policy — reshaping the incentives around future conduct — without any explicit rule change. The law nominally still prohibits these offenses. The pardon apparatus systematically removes their consequences.
The second is foreign-national campaign contributions. Julio Herrera Velutini (Venezuelan-Italian billionaire, Puerto Rico bank owner), Mark Rossini (former FBI agent), and Wanda Vazquez Garced (former Puerto Rico governor) were pardoned January 15, 2026, then amended on January 20 to add "contribution by a foreign national" as an additional covered offense — expanding a pardon after the fact to cover charges the original warrant had not named. Imaad Zuberi's May 2025 commutation, later amended in October, covered FARA violations, tax evasion, and foreign conduit campaign contributions. Ballard Partners lobbyist Brian Ballard's simultaneous roles as expert government witness on FARA violations and paid advocate for Anthropic on Pentagon procurement — documented in Wartime: Lawfare — sit alongside a DOJ under Attorney General Pam Bondi (Ballard Partners partner 2019–2025) that dropped approximately half of its pending foreign bribery investigations in early 2025. Pardoning existing FARA convictions while reducing new FARA prosecutions produces the same policy outcome through two different instruments: foreign money entering U.S. politics becomes practically unenforced.
The third is financial crime. Paul Walczak (tax fraud, $4.4 million restitution) was pardoned April 23, 2025; reporting indicates his mother attended a $1-million-per-plate Mar-a-Lago fundraiser before his pardon. Trevor Milton, the Nikola securities-fraud defendant convicted on approximately $676 million in investor losses, was pardoned March 27, 2025; he is a documented Trump donor. The March 2025 BitMEX cluster — founders Delo, Dwyer, Hayes, and Reed pardoned individually, plus HDR Global Trading Limited pardoned as a corporate entity — represented, per some reporting, the first-ever presidential pardon of a corporate defendant for Bank Secrecy Act violations. Changpeng "CZ" Zhao, founder of Binance, was pardoned October 21, 2025, for the largest crypto-related money-laundering conviction on the books. The crypto pardon cluster occurs while the Trump family's World Liberty Financial is deeply enmeshed in crypto — Zach Witkoff as co-founder, Barron Trump reportedly holding approximately $150 million in crypto and World Liberty Financial profits since the Iran war began, per Wartime: Succession. Pardoning the largest crypto figures for AML violations sends a market signal about enforcement priorities that benefits the family's own crypto positions. The signal is what the piece names; the causation is not asserted.
The fourth is the individual reward pattern that has no historical parallel. Devon Archer, Hunter Biden's business associate, was pardoned March 25, 2025 — after having testified before Congress in 2023 about Hunter Biden's business dealings in ways Republicans characterized as damaging to the Biden family. This is not class indemnification. It is not reintegration. It is a specific individual whose usefulness to the administration's political project is documented in the congressional record, whose federal securities-fraud conviction was then vacated by presidential order. The pardon operates here as reward. Adding it to the modal analysis: the pardon power's operational modes in the current term include mercy (historical case-by-case, still occurring but rare), class indemnification (January 6, FACE Act, 2020 election), reintegration (Kushner, Blagojevich), and reward (Archer). All four operate simultaneously. None of them is the historical use of the power.
Set against these four modes runs a fifth data point that fits none of them: on December 1, 2025, Trump pardoned Juan Orlando Hernández, the former President of Honduras convicted in U.S. federal court of drug trafficking and sentenced to forty-five years in 2024. A U.S. president has never before pardoned a foreign head of state convicted in U.S. courts of narcotics offenses. It is an outlier that demonstrates the pardon power is being used in genuinely novel ways, at a scope the traditional analytic categories do not cover.
Part Four — What This Made Possible, and What Adapts
The two-sided instrument was operational for approximately fifteen months before the Anti-Weaponization Fund's most visible piece — the Fund itself — was rescinded in writing on August 2, 2026. That rescission was extracted by Senators Cornyn and Tillis as a condition of advancing Todd Blanche's confirmation as Attorney General, not by any court. Three lawsuits challenging the Fund — CREW (D.D.C., Leon), Floyd (E.D. Va., Brinkema), and Gill (S.D. Cal., Bencivengo) — remain in various procedural postures; the Floyd preliminary injunction of June 12, 2026 was appealed by the administration on the same day it moved to dismiss the Gill suit as moot, an internal contradiction the government has yet to resolve on any docket. The full posture is documented in the Federal Litigation Index.
What was demonstrated in that window is that the class-protection architecture worked as designed. Pardons issued as class-based proclamations were not judicially challengeable — Trump v. United States protects them absolutely as core constitutional powers, and even the evidence of their exercise cannot be used against the president in any related prosecution. The Fund's design withstood two adverse rulings and one favorable one, and the government's response was to appeal, argue mootness, and refuse to file the written rescission Judge Leon had asked for until a Senate confirmation vote forced it. The three-district split on identical facts — mootness accepted in one court, rejected in another, pending in a third — demonstrates that the underlying instrument is legally live even when individual implementations of it can be paused.
What adapts is more instructive than what was rescinded. The Federal Tort Claims Act route now doing what the Fund would have done — McCloskey's approximately four hundred January 6 defendants filing administrative claims for one to ten million dollars each, Michael Flynn's $1.25 million settlement, the Ashli Babbitt family settlement, Carter Page — routes the same class through a mechanism with entirely different structural properties. Once DOJ agrees to an FTCA settlement, Treasury pays automatically from the Judgment Fund. No separate appropriation. No rulemaking process. No Fund commissioners subject to Senate confirmation. No APA notice-and-comment exposure. The same class of beneficiaries receives compensation for the same category of grievance through a pipe that has no public-visibility exposure comparable to the Fund's.
Two other adaptations sit on the record. The January 20, 2026 amended pardons, which added covered offenses to existing pardon warrants after the fact, extend indemnification retroactively in a form pardon-attorney practice has rarely if ever used. And the January 27, 2025 executive order returning discharged servicemembers with back pay — the class of individuals who had refused the COVID vaccine mandate — accomplishes what the Poffenbarger and Doster military-mandate lawsuits could not achieve through the courts. Class-based compensation delivered by executive order, with no rulemaking process to challenge and no appropriation required to pay it. The mechanism the Fund attempted publicly is achievable through several less-visible instruments. The mechanism can change; the individual beneficiaries can change with it. What remains constant is the political relationship that determines who receives protection. This is not a protected class in the conventional sense. It is a conditional class: individuals enter and leave its protection according to their political relationship to power. The alleged “fraud on the court” is therefore not an isolated procedural defect. It is conduct first immunized through pardon, then converted into a claim for compensation, while the presidential authority behind the arrangement is insulated by immunity doctrine.
The counter-architecture Wartime XX names in Cluster 7 — command responsibility, ICC jurisdiction, universal jurisdiction, EU AI Act product liability, and the record itself — applies here too. What resists the two-sided instrument is the same set of venues that resist the targeting architecture, plus specific accountability proposals emerging in the political layer: Casey Michel's Church Committee framing in Just Security, the three Fund lawsuits, congressional letters from Wyden and Raskin, and the Senate confirmation-vote lever Cornyn and Tillis demonstrated worked once. None of these have yet reached the doctrinal architecture. Trump v. United States remains operative. The pardon apparatus operates continuously.
The instrument
The pardon indemnifies. The Fund compensates. The doctrine protects. Together they form the two-sided instrument the piece is named for — indemnification going backward, compensation going forward, doctrinal insulation from prosecution for having built the arrangement itself. The class the arrangement protects is defined not by objective legal harm but by political relationship to the current administration. That is the categorical shift. The historical pardon power addressed miscarriages of justice case-by-case; this one indemnifies political classes. The Federal Tort Claims Act addresses documented government wrongdoing; the FTCA route now flows compensation to a political class the administration has designated. The immunity doctrine addressed the specific question of whether a former president can be prosecuted for acts as president; it now insulates a general architecture of legal protection whose actors and beneficiaries share a political relationship.
What Cult of Government names in sacralization terms — sacred authority converted into license for exactly the violence its tradition forbade — this piece names in legal-protection terms. Executive clemency and class compensation, historically instruments of mercy and remedy, converted into license for exactly the acts they were meant to correct. Deus Vult tattoo and pardon warrant. The vocabularies differ. The operation is the same. The surface of the instrument — its historical form, its constitutional pedigree, its language of forgiveness and remedy — becomes the protection for its inverted operation.
There is a name for this in constitutional theory too. When a president exercises the pardon power routinely as class-based indemnification for political allies while public prosecutions remain reliably available against political opponents, the office is no longer exercising mercy in the constitutional sense. It is exercising sovereign discrimination — deciding, one class at a time, whose crimes the state prosecutes and whose it forgives. This is not what Article II envisioned. It is what monarchical prerogative looked like before the constitutional check on it. The pardon power was preserved from monarchy specifically to remain narrow — a rarely-used check against injustice, not a routine instrument of class governance. What the current term demonstrates is that the constraint on its scope was always cultural rather than legal. When the culture removes the constraint, the instrument reveals what it always could have done.
The Founders' fear was not that a president would fail to pardon enough people. It was that the power would be used to shield conspirators in offenses against the government itself. That is the specific concern James Madison raised at the Virginia ratifying convention: "There is one security in this case to which the gentlemen may not have adverted: if the President be connected, in any suspicious manner, with any person, and there be grounds to believe he will shelter him, the House of Representatives can impeach him." The check on the pardon power was the impeachment power. In the Trump v. United States framework, both sides of that check now sit inside the executive: the pardon is a core constitutional power protected absolutely; the acts that would have supplied evidence for impeachment or subsequent prosecution are official acts protected by the evidence-exclusion rule. The Madisonian security has been doctrinally removed. What remains is a Congress with the constitutional authority to impeach but without the institutional will or the evidence-access to act. The check exists on paper. The instrument operates.
The Fund is one instrument among several. The pardon is one instrument among several. Trump v. United States is one ruling among several. Read as separate stories, each is defensible on its own terms — a president using his constitutional pardon power, a settlement establishing a compensation mechanism, a Court ruling on a specific immunity question. Read together, they are an architecture. The architecture protects a class. The class defines whose conduct the law now reaches, and whose it does not. That is the instrument this piece is named for. Whether it is dismantled or preserved is the question the counter-architecture — the lawsuits, the congressional oversight, the international jurisdictions, the record — is now attempting to answer. That question is not yet answered. The instrument is still operating.