When Never Equals Maybe
On February 25, 2026, a Homeland Security official got on a call with election officials from around the country and made them a promise. "There will be no ICE presence at polling locations."
It's a reassuring sentence. It's also a strange one. For most of the last century, nobody in the federal government needed to say it. (ICE did say something like it once, in 2018, to counter a fake flyer in Milwaukee. Nobody was asking election officials for a promise.
But to make any of this make sense, we need some background. So... [pause for Google search] Congress made it a crime to have troops at polling places in 1865, at the end of the Civil War, because soldiers had shown up there in the border states. During Reconstruction Congress went the other way for a while and gave presidents the power to send federal marshals and troops to protect Black voters from the Klan. While the net effect was safer elections for minorities, it put the Army in the middle of contested political events, and by 1878 Congress had barred the Army from ordinary domestic law enforcement.
Then, in 1909, Congress took out the exception. The old version allowed troops at the polls to "keep the peace." Congress cut it. The official reason was that no law authorized troops for that purpose anyway. Later pundits pointed out that the phrase could have justified almost anything. What remains today allows armed feds to "repel armed enemies of the United States." Whether or not the ban includes ICE officers is a question that hasn’t been tested in court yet. When the country needed federal protection for voters again during the Civil Rights era, the Voting Rights Act sent observers - without guns.
From what I can find, no president has ordered armed troops to an American polling place in almost 150 years. So the recent promise seems like it should’ve never needed making.
A rule that's doing its job doesn't get announced, because nobody's worried about it. Think about your office building. There's no sign that says "We promise not to lock the fire exits." The fire code covers it, and so does habit. Now imagine the tenants asked the building manager, in writing, whether the exits would stay unlocked this year. The manager's reply doesn't tell you much, but the question does. Something gave the tenants doubt, which created the need for the owner to promise something that hadn’t been a concern before.
That's where DHS was in February. The election officials asked. ICE promised they’d stay away. That should’ve been the end of it.
But then the promise started to shrink. Every follow-up question got a less committed answer in response:
February: "There will be no ICE presence at polling locations."
March, the department head at the time, testifying to the Senate: "There are no plans to have ICE officers at our polling locations." Asked for an absolute commitment, she didn't give one.
May, the acting head of ICE, in a letter to concerned secretaries of state: "there is no reason for ICE personnel to be deployed to polling places."
July, a DHS statement: "ICE is not planning operations targeting polling locations. ICE conducts intelligence-driven targeted enforcement, and if an active public safety threat endangered a polling location, they may be arrested as a result of that targeted enforcement action."
September, the current department head, to reporters: "The only reason why we would be at polling places is if there is a threat to that polling place or we're serving a warrant on someone that we have been actively tracking down."
Read them one at a time and no statement reverses the one before it. Each one just leaves a little more room. Read the first and the last together and they contradict each other. February said no presence. September said presence, under conditions. Nobody ever announced a change in position, because the position never changed in one move. It moved in five.
I have no inside knowledge of how DHS works or what it's planning. I'm reading public statements, and three explanations fit them equally well.
It could be deliberate. Leaving the answer open keeps every option open.
It could be legal caution. There have always been threats of some sort on election day, and legitimate threats are cause for valid warrants. Maybe the February promise was the mistake and September is the accurate version.
It could be that nobody owns the decision. Each official describes the rule their own way, and the description drifts toward the thing least difficult to violate.
I can't tell you which one it is, and neither can anyone else out of the loop. But there’s stuff that doesn’t depend on which explanation is right.
Looking at this using Linden's Bridge (the organizational theory this newsletter is built on), the problem is in the Rules layer. Rules is where an organization determines who gets to decide. For 117 years the written rule and the operating rule matched: no armed federal agents at the polls. The September version adds an exception, "a threat or a warrant," and an exception is only as stable as its owner and its criteria.
Right now the answer to both is "nobody has said."
Let’s start with the warrant. There are two kinds. A judicial warrant is signed by a judge, someone outside the agency who can say no. An administrative warrant is a form signed by a supervising officer inside the agency, and nobody outside the executive branch reviews it. It also may not let agents into some areas without consent, which matters at a polling place. If the exception runs on judicial warrants, a judge owns a piece of the rule. If it runs on administrative warrants, the department owns all of it, and the "warrant" limit is the department promising itself it'll be good. The September statement didn't specify which.
We can examine the threat, too. Nobody has published who decides what counts as one. California's secretary of state asked about advance notice on that February call, and the answer she got was that the whole idea was "disinformation." I couldn't find any answer since then, under the newer exceptions.
When an exception has no outside or visible owner, the decision drops to whoever is standing there. On November 3, and at early-voting sites, that's going to be an agent in the field, at one of nearly 100,000 polling places, who has heard the department head say they're allowed there if there's a threat. As of today, nobody has told that agent what a threat is.
Why the usual fixes haven't worked
The obvious answer is a list of things the department could do. Name an owner. Define the test. Seven months in, neither has been announced, and any fix that relies on the goodwill of the party being constrained is a fix that ain’t gonna hold. That's about the oldest finding in my work.
So the states and the plaintiffs have gone around the department. At least seven states passed laws this year to keep armed agents away from polls. Two lawsuits were filed in September asking a federal judge to bar them. The question has been put to officials in congressional hearings more than once. So far none of it has changed the department's stated position.
I’m not complaining here. It’s something to take seriously, because the frameworks I’ve developed get specific about why this is happening, and the lessons apply to any administration, party, or organization that finds itself here.
Start with the rule. A rule with an exception can only stay stable, regardless of who is in office, when three things are true.
The test for the exception sits outside the control of the group that can apply it.
The consequence for breaking it lands on the person who acts.
Specific questions about it get specific answers on the record.
Each moves a piece of the exception out of the department's hands. A judge decides what a warrant is. A prosecutor decides whether to charge the agent. A hearing decides what "threat" means. That's the trade, and there's no version of it that comes free. A rule that's stable for the system is a rule that's less flexible for whoever is running the system right now. This is true for any administration. It's the choice between stability for the structure and power for the party in the building, and the department's answers have moved toward more room every month since February, one sentence at a time.
I don't say that as an accusation. Every administration I can think of has wanted more room than the rules give it. The question the frameworks ask is what that costs, and who ends up paying.
The consequence, and the layer where it lands
Linden's Bridge has a layer called Execution. It's the layer where work actually happens, as opposed to where it's decided. A department head doesn't walk into a polling place. The President doesn't serve a warrant. The decision about the exception sits in Rules, but the act happens in Execution, by one agent, in one place, with his or her name on a form.
Most of the response so far has aimed at the institution. Sue the department, or ban it. An institution can absorb that. It appeals in the courts and outlasts the story. The federal law aims lower. Up to five years in prison for the officer who "orders, brings, keeps, or has under his authority or control" armed men at the polls. It went after the person who did the thing. As far as I can tell, that personal exposure still exists in federal law today. The structural question is whether anyone with the power to enforce it will.
When the federal government won’t act, state law may be able to fill the gap. A state can't fence federal agents out of a polling place. A federal officer doing their lawful job is immune from state rules. Several of the states that passed laws this year wrote the federal ban into their own criminal code. The theory is that an officer who brings armed men to a polling place may have stepped outside any lawful federal job, and if he has, the immunity doesn't cover him. A state prosecutor could then charge him under the state's copy of the law.
That's a real fight, but the states aren't guaranteed to win it. An officer can move the case to federal court, where the government would likely claim he believed he was doing his job. Whether the 1865 law even covers two agents serving a warrant isn’t clear. What the state route does add is a thing the federal government can't remove: a state conviction can't be pardoned by the President. It's the only consequence on the board that doesn't run through the executive branch, and it lands on an individual, not an agency. Whether it ever gets used depends on a state prosecutor and a judge.
The test, and why open-ended statements get more specific when questioned
Linden's Lens looks at three things separately: what an organization believes, what it says, and what it does. An open-ended statement leaves every option available on the doing side, whether or not anyone chose it that way. That's true of "a threat or a warrant" right now.
But the statement narrowed under questioning. Four times. Every direct question from a senator, a state official, or a reporter got a bit more specific response. So... what the department says can change. It just depends upon what’s being asked. "Will ICE be at the polls?" has been asked over and over. "Will states get a call first?" was asked once, in February, and dismissed. "What counts as a threat?" and "which warrant?" haven't been put to the department on the record at all, as far as I can find. And the two lawsuits that could force answers were only filed in September. A ruling on a request to block deployment could come before November 3. Depositions and discovery of the field guidance won't. This part belongs mostly to the courts and to Congress, and the specific questions are still sitting there, unasked.
Why the pace is so slow
Earlier this month I ran DHS’s public behavior through Loadmap's archetype read, which sorts organizations by how they move under load. The pattern I found (without a ton of confidence - I’m not on the inside) is one Loadmap calls Special Ops: fast, act-first, keeps information close, does things its own way. It's a guess. But the pattern has a known weakness: a system that answers everything fresh, with nothing pre-built, can’t withstand sustained pressure from several directions at once.
Think about a literal Special Ops team. They pick their target, choose the moment, execute, and improvise when conditions change. What they don’t do is stand out in the open against multi-front pressure. That’s a circumstance that ends teams like this.
The responses so far have been the kind of things a special ops team can handle. It’s all coming one at a time and it’s stuff they can prepare for. A proposed law, a big case, a hearing, each on its own calendar. Against a fast-moving agency, a single large case is the easiest thing to handle, because it has one front and a schedule. Pressure from all sides is different: prosecutors in several states, each independent... records requests to each field office... specific questions in every forum that can force an answer. That's what makes an external check effective against a fast actor, and it would be just as true if the other party was in power.
Where this leaves things
No one outside the hierarchy can actually force a change in its stated policy. What outsiders can do is make the three conditions true without the department's help: put the test in a court's hands, keep the consequence on the person who acts, and require specific answers on the record. None of that depends on anyone's intentions or on a fast federal court. It does depend on state prosecutors, judges, and people willing to keep asking the same specific question, which is where the load on the department is most likely to force a change.
Here's what the trade looks like from inside the department, as best I can read it. The vague version gives the department freedom now. It also depends on nobody else acting, and other people have already started to act: seven state laws, two lawsuits, a hearing schedule. Every one of those is a response to the vagueness, and each costs the department time and attention it didn't want to spend. Power that rests on other people staying out of it lasts as long as they do.
The written version costs the department some freedom and gives it something the vague version can't. A rule the department wrote down and owns is a rule the other side can plan around, which means fewer surprises and fewer suits. It also binds whoever comes next. That's the part I'd want a party in power to think through. The law from 1865 limited the party that passed it, and then it limited everyone who came after, including the opposition. The same is true of 1909 and 1965. An exception with an external owner protects you from the next administration exactly as much as it limits you now. The only way to get that protection is to accept the limit first.
That's why this isn't about one administration. A rule with an external owner and a personal consequence works the same on everyone. A rule with no owner works for whoever is using it, right up to the moment someone else acts.
Back to organizations
You've lived a version of this. "There will be no layoffs." Then, a quarter later: "There are no layoffs planned." Then: "Any changes will be performance-based." Then: "We're reviewing the organizational structure." Nobody lied. Every sentence might be accurate. And the staff started checking the fire doors at step three.
What didn't happen in that story? Nobody asked "which roles?" Everyone heard the shrinking answers and started updating their resumes. That's the normal response, and it's the one that lets the promise keep shrinking. A vague answer only stays vague while the questioners can’t demand specifics.
When a promise you used to take for granted suddenly needs saying out loud, don't argue with the promise. Ask the specific question, on the record, and keep asking until you get a specific answer. "Which warrant?" "Which roles?" "Who signs off?" Each specific answer takes an option off the table, and it does that whether or not the person answering wanted it to. That's what worked on the department four times this year. It works on a CEO too.
If you want to apply it to November, it's cheap. Your county clerk and your secretary of state know whether anyone has told them what "threat" means and whether they'll get a phone call first. Ask them. Ask your representative to ask. Every specific question so far has moved the department's answer. What it hasn't been asked, it hasn't had to answer.
Never became maybe in five sentences. Make them say which one they mean.
