Skip to main content

Politicians make promises on their stump — we watch and hold them accountable.

Help keep the record honest →Create an account

Issue

Guns & the Second Amendment

Firearm regulation and firearm rights, including background checks and the Second Amendment.

In short

The modern Second Amendment is very new and almost entirely the Supreme Court's work: seven decisions between 2008 and 2026 draw the boundary, not any statute. Heller recognised an individual right. McDonald applied it to the States, though the passage most often quoted from it is a four-Justice plurality rather than an opinion of the Court. Caetano vacated a state ruling for contradicting Heller's holding that "arms" reaches weapons that did not exist in 1791, while stopping short of holding stun guns protected. Bruen replaced means-end balancing with a history-and-tradition test and reserved the 43 States' shall-issue licensing regimes in the opinion of the Court at footnote 9, not merely in a concurrence. Rahimi loosened that test, upheld temporary disarmament following a judicial finding about a specific person, and rejected disarming someone simply for not being "responsible," which it called a vague term. In June 2026 the Court decided two more: Hemani, holding a prosecution based solely on admitted marijuana use inconsistent with the Second Amendment while saying in terms that the case is narrow, and Wolford, reversing the Ninth Circuit six to three over Hawaii's rule on private property open to the public. On June 30, 2026 the Court granted certiorari on whether the Second and Fourteenth Amendments guarantee the right to possess AR-15 platform and similar semiautomatic rifles. That is open, as is whether a nonviolent conviction may disarm a person for life. StumpWatch's position, stated as a position and not as law, is that a private party carrying out a government firearms programme is a government agent, so the test is whether the same act would violate the Constitution if an official performed it directly. Vullo states that principle unanimously, as a bar on doing indirectly what an official may not do directly, but as First Amendment law about advocacy; no court has extended it to arms. Four things a reader would not expect are on the record: the National Firearms Act transfer and making taxes are now zero for everything but machineguns and destructive devices, and a district court has held the registry resting on it exceeds Congress's enumerated powers; the 2024 dealer rule was vacated; the definition of an unlawful drug user was narrowed by an interim final rule, operative now but not yet final, so occasional use no longer triggers the disability; and section 925(c) promised prohibited persons a route back while appropriations riders defunded it for individuals, with a clause of the same proviso keeping it open for corporations. The websites of ATF, GAO and the Centers for Disease Control refuse automated retrieval, and the categories of evidence those three hold are the ones that would most test this page's claims, so the gap is a real limit on the analysis and not merely a limit on its routes. Two figures did get through, and by one route rather than two: Congressional Research Service In Focus IF12057 carries both sentences, ATF's count of nearly 921 million out-of-business records as reported to a congressional inquiry, and GAO's 2016 finding that ATF's out-of-business imaging system complied with the registry prohibitions as the Congressional Research Service describes it. No ATF or GAO document was opened here. Both are on the page above, and the GAO finding cuts against this page's argument. What remains genuinely absent is ATF's trace, commerce and compliance-inspection data, GAO-16-552's own text, including the partial-non-compliance findings an earlier version of this page quoted and this revision could not verify, and the Centers for Disease Control's injury and mortality data. The Supreme Court's site, the FBI and the Congressional Research Service all answered on first request when they were tested on August 11, 2026, and an earlier version of this page wrongly said they refused.

Explore the record

Everything behind this page sits in the 6 sections below — the claims, who is on record, what the public thinks, and every citation. Choose one.

Constitutional analysis

This is our assessment of the issue against the Constitution — not a sourced fact and not a court holding. It separates what the law establishes today, what is genuinely contested now, and the argument for where the law should go. Written and adversarially reviewed away from this site, then stored; the page renders it and never generates it. Everything behind the tabs below is that assessment, it is open to dispute, and the counter-arguments that survived review are published in full under Summary.

What does the Second Amendment actually protect today, and may the government accomplish through private hands what it is forbidden to do itself?

Key points

  • The modern Second Amendment is recent and almost entirely the work of the Supreme Court: seven decisions between 2008 and 2026 draw the boundary, not any statute.
  • In June 2026 the Court decided two cases seven days apart. Hemani held a prosecution based solely on admitted marijuana use inconsistent with the Second Amendment, and said in terms that the case is narrow. Wolford reversed the Ninth Circuit six to three over Hawaii's rule flipping the common-law default on private property open to the public.
  • On June 30, 2026 the Court granted certiorari in Viramontes and Grant on whether the Second and Fourteenth Amendments guarantee the right to possess AR-15 platform and similar semiautomatic rifles. It is undecided.
  • Whether a nonviolent conviction can cost a person the right for life is unsettled. The Sixth Circuit holds section 922(g)(1) valid on its face and as applied to dangerous people, while requiring that an individual get a real chance to show he is not dangerous.
  • StumpWatch's position is that a private party carrying out a government firearms programme is a government agent, so the test is whether the same act would violate the Constitution if an official did it directly.
  • Courts have not adopted that reasoning on these facts. Vullo states the anti-circumvention principle unanimously, as a bar on doing indirectly what an official may not do directly, but as First Amendment law about advocacy rather than a Second Amendment holding about arms.
  • Congress set the National Firearms Act transfer and making taxes to zero in 2025 for everything except machineguns and destructive devices. In August 2026 a federal district court held that the registration provisions could no longer rest on the taxing power and enjoined them as to untaxed firearms for the parties before it.
  • The guarantee against a national firearms registry is entirely statutory. It rests on a limit on rulemaking, two appropriations provisions, a clause of the Brady Act and a destruction requirement in the Gun Control Act. None of them is a constitutional holding and any of them can be repealed by ordinary legislation.
  • Federal law told prohibited individuals for three decades that they could apply for restoration while appropriations riders defunded the individual process, and the Supreme Court held in Bean that the resulting inaction produced no reviewable denial. A clause of the same proviso keeps the process open for corporations. It is absent from the original 1992 rider and present in every later act this page opened, at fiscal years 2012, 2024 and 2026; the years in between were not opened, so this page does not say when it began.
  • Even a constitutional win may produce no remedy. Vullo won unanimously on the pleadings and got a remand with qualified immunity still open; Hemani won unanimously in the judgment and the statute is untouched.

What is at stake

WHAT IS AT STAKE

Twenty-seven words: "A well regulated Militia, being necessary to the security of a free State, the right of the people to keep and bear Arms, shall not be infringed." For most of American history the Supreme Court said very little about them, and the decision most often cited from that long silence is United States v. Miller, decided in 1939. Since 2008 the Court has said a great deal, and the outer boundary of what any government in this country may do about firearms is now drawn by a short line of decisions rather than by any statute: Heller in 2008, McDonald in 2010, Caetano in 2016, Bruen in 2022, Rahimi in 2024, and two more handed down in June 2026. Seven decisions, in eighteen years. Two questions sit underneath every fight on this issue, and they are not the same question. The first is what the right covers. Which arms, carried where, by whom, and on what showing may the government take them away. That question is being answered right now, one weapon and one prohibited category at a time, by a method the Supreme Court adopted in 2022 and has been adjusting ever since. A great deal of it is genuinely open, including whether an entire category of commonly owned rifles is protected at all. The Court agreed on June 30, 2026 to decide that. The second is whether the government may accomplish through private hands what it is forbidden to do itself. That question gets far less attention and it is the one StumpWatch treats as decisive, because much of the operating machinery of American gun regulation runs through private parties. A federally licensed dealer, not the government, runs the background check and creates and keeps the record of who bought what. A property owner, not the government, decides in some States whether a permit holder may cross a threshold armed. And in one documented case a state regulator urged banks and insurers to cut ties with a firearms advocacy organisation, which produced a unanimous Supreme Court decision discussed further down this page. Those three mechanisms are not equally well established and this page does not treat them as if they were: the first two are statutory and structural, the third rests on allegations a court took as true at the pleading stage. StumpWatch's position, stated as a position and not as law, is that a private party carrying out the government's programme is an agent of the government, and that constitutional protections follow the conduct rather than the badge. The test we apply to every issue on this site is the same one here: would this be a constitutional violation if a government official did it directly? If it would, it is the same violation when an authorized agent of the government does it instead. One thing should be said plainly at the outset, because it is where this issue differs from most. StumpWatch's editorial lens favours rights and civil liberties, and on guns that lens does not point in one partisan direction. It points toward a robust Second Amendment. It also points toward due process for the person the state wants to disarm, toward the nonviolent offender carrying a lifetime disability for a decades-old conviction, toward the state-licensed cannabis user who commits a federal felony by owning a shotgun, and against a government that regulates by pressuring the banks rather than by passing a law. Those commitments do not always agree with each other, and where they collide this page says so rather than picking whichever answer is convenient. What follows separates three things and keeps them separate: what courts have actually held, what is genuinely open, and what we argue the law should be. The third is ours. It is not law, it is labelled, and no ruling changes it.

Counter-arguments that survived review

These were argued against the analysis at full strength and were not defeated. They are published here, unfolded, because an assessment that hides its strongest surviving objection is advocacy rather than accountability.

STATUS OF THIS ANALYSIS. It has now been through research, primary-source verification and six independent adversarial passes: constitutional doctrine, statutory and regulatory, empirical and neutrality, a verbatim-quotation audit, and the two earlier lenses on citation and omission and on state action and remedy. That meets the Issue Analysis Standard's minimum of five. It is nonetheless stored as a DRAFT, for a reason given below. THE FIRST THING TO SAY IS THE REASSURING ONE, AND IT WAS ESTABLISHED THREE TIMES INDEPENDENTLY. There are no fabrications on this page. No invented case, no invented docket, no invented quotation. Sixty-two quoted passages were checked character by character against the primary document and verified exact. That number belongs to the verbatim audit and to the text as it stood when the audit ran; the corrected text carries roughly two hundred quoted spans, and no audit of that scale has been run over it, so the figure should be read as what was checked then rather than as a measure of the page as it now stands. The sourcing work behind this analysis was real. What the passes found instead was a large number of citation, attribution, provision-status and framing defects, and one class of defect nobody had looked for. THE CLASS NOBODY HAD LOOKED FOR. Verification had asked whether a statute says what this page says it says. It had never asked whether the provision was still operative. Four separate statements on this page were accurate about the text and wrong about the law: the National Firearms Act transfer and making taxes were described as live when Congress set both to zero in July 2025 for everything but machineguns and destructive devices; the section 5841 registry was described as simply existing when a federal court had held in August 2026 that it exceeds Congress's enumerated powers and enjoined it as to untaxed firearms for the parties before it; the 2024 dealer rule was cited as live law after being vacated in June 2026; and the Form 4473 discussion described a trap that a January 2026 interim final rule, operative but not yet final, had already narrowed so that it no longer catches an occasional user. Any future analysis on this site should treat provision status as a distinct verification lane rather than a by-product of reading the text. THREE EXCUSES THAT WERE NOT TRUE. An earlier version of this page told readers that the Supreme Court's website, fbi.gov and the Congressional Research Service all refused automated retrieval, and cut material on that basis in at least eight places. All three answered on first request. This is a worse category of error than a bad pin cite, because a reader cannot detect it by reading the page, and because it converts an unfinished search into an apparent property of the world. The material has been restored: the June 30 2026 grant order and the question presented are now quoted from the Court's own docket, the separate writings at the denial in Snope are back, the VanDerStok citation is given, the FBI's background-check totals are on the page, and the four-provision registry framework has been credited to the Congressional Research Service product it was taken from. What genuinely refuses, retested and reproduced, is atf.gov, gao.gov, cdc.gov and stacks.cdc.gov, all returning HTTP 403. ONE CORRECTION RUNS AGAINST THE PREVIOUS CORRECTION. An earlier version of this page said GAO-16-552 could not be retrieved, then said that statement was false and quoted six passages from the report. This revision could not retrieve it either: gao.gov returns HTTP 403 and the copy on govinfo.gov returns an HTTP 200 status with an error page as the body, which is precisely how a document comes to be quoted by someone who never opened it. Those six quoted strings are marked UNVERIFIABLE rather than repeated or silently dropped. The report's most adverse finding for this page's argument is retained because an independent primary source, the Congressional Research Service, describes it. WHY NO CONFIDENCE FIGURE IS PUBLISHED. The verdict split is published instead, as raw counts, and can be read directly: of 160 load-bearing claims, 100 survived the six passes as stated, 25 were overstated and have been narrowed, and 35 did not survive as stated and were withdrawn, inverted or replaced. Converting that ratio into a confidence would misdescribe what it measures. It measures the SUPERSEDED text, because that is the text those reviewers read. Two further adversarial rounds have since run over the corrections themselves, but they do not convert into a confidence either: they returned a shrinking defect count together with an unresolved finding that the corrections lean in this page's own direction, and the fixes made in answer to the most recent of them have not themselves been reviewed. A number attached to any of that would be invented, and the Standard says a confidence that cannot be defended is published as absent rather than as a number. WHAT REMAINS OWED BEFORE THIS MAY BE PUBLISHED. A further adversarial pass. Two have now run over the corrections, and the most recent found that their residual drift still favoured this platform; the fixes made in answer to it are themselves unreviewed, and the directional finding recorded below is not resolved. And a rewritten issue summary bound to the new body hash. WHAT FOLLOWS ARE THE OBJECTIONS THAT SURVIVED. They are not defeated. VULLO IS A SPEECH CASE AND THE EXTENSION IS OURS. Vullo holds that officials may not coerce intermediaries to punish disfavoured speech. Reading it to govern a State's use of banks, processors or dealers to burden the acquisition of arms is an extension no court has made, and the Court decided the case on the First Amendment precisely because the NRA's injury was to advocacy. This revision found that the opinion of the Court states the principle more broadly than the syllabus does, as a bar on doing indirectly what an official may not do directly, and that formulation helps us. It does not answer the objection. A general-sounding sentence in a First Amendment case is still a First Amendment case. WOLFORD DOES NOT VINDICATE THE AGENCY TEST, AND IS WEAKER FOR US THAN IT LOOKS. The Court held Hawaii's rule cannot stand as a burden on the right and remanded. It said nothing about state action. Worse for our reading, Hawaii criminalised the permit holder's conduct directly and imposed no duty on the property owner, so it is not even a clean instance of a State acting through a private hand. The structural resemblance to our argument is real; the doctrinal support is not. HELLER'S DICTA CUT DIRECTLY AGAINST THE DEPUTISED-DEALER ARGUMENT. Heller treats conditions and qualifications on the commercial sale of arms as presumptively lawful. Dealer licensing, recordkeeping and background checks are the paradigm of such a condition, and eighteen years of decisions have relied on that sentence. Our answer is that the analysis should follow function rather than ownership, but no court has accepted it and the objection is not answered by restating our test. THE FUNCTIONAL TEST MAY PROVE TOO MUCH. If a licensed business complying with a federal recordkeeping mandate is a constitutional actor, the same reasoning reaches every regulated industry that keeps records at the government's direction. We say our version is narrower because it turns on the government running the programme. A reviewer may reasonably say the line we draw is not principled enough to hold, and Halleck's traditional-and-exclusive-public-function exception, which we now quote, has never been read as broadly as we would need it read. THREE FACTS ON THIS PAGE UNDERCUT THE REGISTRY ARGUMENT AND ARE NOT ANSWERED. The background-check system is required by statute to destroy its records of a cleared purchase. The federal archive of out-of-business dealer records consists, on the Congressional Research Service's account, of non-searchable images accessible only for tracing. And the FBI's own note on its background-check totals states that they do not represent firearms sold and that no one-to-one correlation with sales can be made, which limits what the volume figures we now cite can be used to show. All three are stated above rather than buried, and none is refuted. A reviewer could fairly say that once they are on the table, pointing at the gap between popular understanding and statutory text is rhetoric rather than analysis. HEMANI IS NARROWER THAN THIS PAGE'S USE OF IT. The Court itself wrote that in many respects the case is a narrow one, and reserved addicts, the presently intoxicated, individualised prosecutions, future prophylactic statutes and section 922(g)(1) by name. The judgment was unanimous but the Court produced four separate writings. Any reading of Hemani as a general rule against status-based disarmament is an inference, not a holding. THE PROCESS ARGUMENT AGAINST EX PARTE ORDERS IS WEAKER THAN THE DRAFT MADE IT LOOK, AND WEAKER AGAIN AFTER THIS REVISION. The one state statute examined here requires a hearing within fourteen days, requires clear and convincing evidence for a final order, caps that order at twelve months, and gives the respondent a right to move to vacate. This revision also withdrew the claim that the statute was representative: it restricts petitioners to law enforcement, which many States do not, so it is atypical on the feature the argument turns on. No court has held that an ex parte risk protection order violates due process on this ground, and this page has surveyed no other State. THE CIRCUIT ALIGNMENT ON FELON DISARMAMENT IS STILL NOT MAPPED HERE. One court of appeals decision was verified in its own text and the rest were removed. A reader cannot learn from this page how the courts of appeals currently divide on as-applied challenges to section 922(g)(1). The excuse previously given for that gap, that the Court's docket could not be retrieved, was false; the honest statement is that the work was not done. THE STATE-ACTION CANON WAS NOT RE-READ IN FULL, AND A REVIEWER NAMED OMISSIONS. Halleck and Bean were read in the United States Reports for this revision, but Rendell-Baker, Blum, Skinner and Lugar are carried forward from earlier work. A reviewer identified further authorities bearing on the question that this page does not address at all, including Jackson v. Metropolitan Edison Co., Flagg Bros. v. Brooks, American Manufacturers Mutual Insurance Co. v. Sullivan, and Printz v. United States. Their content is not asserted here because it was not verified here, and their absence is a real gap rather than a stylistic one. THE EMPIRICAL CASE IS STILL ASYMMETRIC, AND NOW HONESTLY SO. The opposing position on this issue is engaged doctrinally throughout this page and barely engaged empirically. That asymmetry was previously excused as a uniform constraint imposed by sources that refused retrieval, which was untrue for three of the four sources named. Figures from the FBI and the Congressional Research Service are now on the page. But ATF, GAO and the Centers for Disease Control genuinely do refuse, and the categories of evidence they hold, trace data, compliance-inspection outcomes and injury statistics, are the ones that would most test this page's claims about how the system works in practice. The asymmetry is therefore stated as an asymmetry rather than as a constraint, and it remains a real limit on what this analysis can claim. THE DEALER-RULE VACATUR MAY NOT MEAN WHAT THIS PAGE SAYS IT MEANS. The judgment vacating the 2024 dealer rule says on its face that the defendants may not apply it to anyone. The government moved on July 10, 2026 to amend that judgment to limit the vacatur, and the docket retrieved for this revision shows no ruling on that motion and two entries missing from the public archive. This page reports the judgment and the pending motion together, and a reader should treat the scope as unsettled. THE NATIONAL FIREARMS ACT RULING IS ONE DISTRICT COURT, AND ITS STAY RAN ONLY TO AUGUST 12, 2026. The judgment entered August 5, 2026 disposes of two consolidated cases, Silencer Shop Foundation v. ATF and Jensen v. ATF, and it is a single district-court judgment: party-scoped, reaching only untaxed firearms, decided on enumerated-powers grounds with the Second Amendment claims dismissed as abandoned, enjoining two further provisions as to the Jensen plaintiffs alone and expressly not reaching the Act's regulation of "any other weapon" as to them. The court stayed the effect of its judgment for seven days from entry, so the stay ran only to August 12, 2026. Whether a notice of appeal was filed or the stay extended could not be determined when both dockets and the Fifth Circuit's records were checked again on August 11, 2026; the lead docket's public mirror ends at the final judgment, the Jensen docket shows the case terminated on consolidation in June 2026 with nothing later, and the appellate archive lists no docket for the case. That is an absence of evidence rather than evidence of absence, the reading it invites would favour this page, and this page labels the appellate position unverified rather than asserting that no appeal was taken. Anyone reading this after August 12, 2026 should check the docket before relying on the description. It is reported because a page that told readers a federal registry exists must tell them when a court has held that Congress lacked the power to create it, but it is not the law of the country and nothing on this page treats it as such. THE HIDDEN AGENDAS SECTION RUNS ONLY ONE WAY, AND THAT FINDING STANDS UNANSWERED. Every item in that section examines a regulator or a legislature acting to restrict, so the section's method only ever runs in the direction of this platform's own editorial lens. A section that checks one side is not applying a method to the record; it is decorating a conclusion with one. A search was made for an item running the other way and it did not produce one that meets the section's own schema. The best candidate, the 2025 repeal of the National Firearms Act transfer and making taxes alongside a registry that survived it, is on this page but sits with the statutes, because there was no documented statement for it to diverge from. The section states all of this about itself. Stating an objection is not answering it, and it is recorded here as well, because an objection a page knows about and keeps out of its published objections is worse than one it never found. THE CORRECTIONS TO THIS PAGE HAVE THEMSELVES LEANED IN ONE DIRECTION. The corrections were put through two further adversarial rounds. The second round found fewer defects than the first, which is what convergence looks like, and then found something a defect count does not capture: the residual drift still ran toward this platform's position, with several separate corrections each shading in its favour. Fixing accuracy had not fixed direction. The mechanism is ordinary and does not require anyone to write a falsehood, which is why it survives review that hunts for falsehoods: an author who knows what a page argues tends to resolve each ambiguous phrasing the same way, and the individual choices all look defensible. That is a defect in the process that produced this text as much as in the text, and no pass has yet shown it corrected. A reader should assume that close calls in the wording here are more likely to favour this page's argument than a neutral drafter would have made them, and should weight the raw claim counts published above more heavily than the confidence any given sentence projects. OUR OWN LENS CONTAINS AN UNRESOLVED CONFLICT. A commitment to rights and civil liberties points toward both a robust Second Amendment and robust due process for the person the state seeks to disarm, and offers no rule for the case where arming one person endangers another. This page states a starting position rather than resolving that conflict, and a reader who thinks the conflict is the whole question will find this analysis incomplete.

How this was reviewed

Assessment · AI analysis · confidence not established
Claims assessed
160
Survived review
100 of 160
Adversarial passes
6
Confidence
Not recordedThe review did not yield a defensible number, so none is shown.

Version 2 · published 2026-08-13 · every earlier version is retained, never deleted.

StumpWatch is live, and the record is still growing. Many promises and positions aren’t tracked yet, and some features are still in beta. Add a sourced promise and help keep the record honest.

Help keep the record honest →