I keep coming back to the same question whenever a national election draws close. If the machines, the databases, and the polling sites are supposed to be the quiet backbone of a vote, why does every cycle arrive with a new stack of warnings? This week the country’s cybersecurity agency put a fresh election infrastructure security plan on the table, just weeks before voters head back to the polls on November 3. It is only thirteen pages. That shortness is part of the point. Officials wanted something usable, not another binder that sits unread until after the last precinct closes.
What The New Election Infrastructure Plan Actually Tries To Do
The document is less a thriller and more a field guide. It maps cyber and physical risks to voting systems, then lists no-cost services the agency can offer state and local officials. Homeland security leadership asked for the plan in July and wanted it out by mid-August. It arrived later than that first target, which already tells you how messy coordination can get when dozens of jurisdictions are involved.
The tone is blunt. Election security is framed as national security, not a talking point for one party or the other. I’ve found that line lands differently depending on who is reading it. Some people hear reassurance. Others hear a reminder that trust has been thin for years and nobody gets to pretend otherwise.
Election security is not a partisan issue. Election security is national security.
That sentence is the spine of the whole packet. Everything else is a checklist of how systems fail, how people fail, and how old-fashioned process can still catch both.
The Threat Picture Officials Want Counties To Take Seriously
The plan does not name a single 2026 plot tied to Russia, Iran, or China. That absence surprised me at first. Then it made sense. A public playbook that names one adversary in detail becomes a political football by lunchtime. Instead the text talks in categories that travel well across counties of every size.
Software flaws sit near the top. Voting equipment is still software sitting on hardware, and software ages poorly. A forgotten patch, a vendor delay, a configuration that nobody documented after a clerk retired in March—those are the boring failures that actually move numbers. Voter registration databases come next. Those files are juicy because they are large, they are updated constantly, and they sit at the junction of state IT shops and local offices that may still run on shared passwords and leftover laptops.
Insider risk is treated with unusual honesty. Not every problem arrives through a foreign keyboard. Sometimes it is a tired worker, a contractor with too much access, or someone who thinks a shortcut will save an hour on Election Night. Physical risk is in the mix as well. Bomb threats against polling places in 2024 are cited as the kind of disruption that does not have to succeed in order to work. Fear empties a line. A closed site pushes voters to a different location. Confusion does the rest.
- Software defects and unpatched voting equipment
- Intrusions against voter registration databases
- Insider access that bypasses normal checks
- Physical intimidation and bomb threats at polling sites
- Supply-chain exposure before a machine ever reaches a warehouse
None of that is science fiction. It is the same family of problems hospitals and power utilities already live with, except the public measures success in a single day and then argues about it for months.
Why Old Procedures Still Matter More Than Fancy Tools
Here is the part I wish more coverage would sit with. The plan keeps returning to practices that sound almost quaint. Ballots handled in bipartisan pairs. Observers in the room while ballots are counted. Chain-of-custody logs that follow a box from the precinct to the canvass. Those habits were not invented for television. They were designed to shrink the blast radius of an insider problem.
In my experience, people glaze over when they hear chain of custody. It sounds like warehouse talk. It is actually the difference between a rumor and a reconstructable event. If two people from different parties move a sealed container, sign the seal number, and photograph the transfer, a later claim has somewhere to land. If one person tosses a bag in a trunk and hopes for the best, you are stuck with vibes.
Many longstanding election practices such as handling ballots in bipartisan teams of two, allowing observers to be present during ballot counting, and maintaining chain-of-custody procedures were specifically designed to reduce the likelihood and impact of insider threats.
Paper is the other quiet star. The plan urges jurisdictions to use paper ballots that can be reviewed, audited, and counted again if software acts up. That is not nostalgia for graphite. It is an admission that code fails and that a voter-marked sheet is still the simplest backup copy the public understands.
I have sat through enough briefings to know the counterargument. Paper is slow. Paper needs storage. Paper needs people who can still read a mark that looks like a check and a scribble at the same time. All true. The plan’s bet is that speed without a paper trail is a bad trade when the argument after Election Night is about whether the count can be shown, not merely asserted.
China-Made Components And The Supply Chain Nobody Wanted To Discuss
A separate letter from the homeland security secretary, circulated in late August, put a sharper edge on the hardware question. Parts inside many machines used across the country are manufactured in China. That fact is not new to procurement officers. It is newly uncomfortable in a public letter.
The concern is not a cartoon chip that flips a race by itself. It is the idea that a component can be altered early, before assembly, before testing, before a county clerk ever sees the box. If that happens, a later network scan may never catch it. You are hunting a problem that was born on a factory floor, not in a phishing email sent the week before early voting.
Voting machines with components manufactured by foreign adversaries introduce significant security risks. These vulnerabilities mean that malicious actors could potentially compromise election infrastructure at any stage of the supply chain, even before the machines are assembled or deployed.
Perhaps the most interesting aspect is the phrase synchronized attacks. That is the nightmare version: not one machine in one gymnasium, but a class of devices sharing a tainted part. Even if that scenario never materializes, the procurement lesson is the same. Counties buy what vendors sell, vendors buy what global plants make, and the public inherits the whole chain.
I do not think every board of elections can rip and replace hardware before November. Budgets do not work that way. Certification does not work that way. What they can do is inventory, isolate, log, and keep paper as the referee. That is less glamorous than a new machine unveiling. It is also more realistic.
Voter Data, Foreign Targeting, And A Wider Definition Of Interference
In July, senior intelligence and national security officials publicly backed a narrower but important claim: China and affiliated actors acquired voter registration information covering a huge slice of the adult population, including some fields that were not sitting in public lookup tools. A White House transparency fact sheet put the figure as high as 220 million records. Treat that number with the usual caution that large estimates deserve. The direction of the finding still matters.
Officials also restated a definition that a lot of commentary skips. Interference does not require a changed tally. Targeting registration infrastructure or the data inside it can meet the threshold on its own. That is a policy statement as much as a technical one. It means a breach that never touches a tabulator can still be treated as an attack on the election environment.
Why steal registration files? Because they are a map. Names, addresses, districts, history of participation. Combined with other leaks, they help someone craft better phishing, better intimidation, better rumors about who is or is not on the roll. They also help identify which local systems are sloppy. A noisy database is a billboard.
| Risk Area | What Breaks First | Practical Counter |
| Registration systems | Account takeover and data theft | MFA, logging, least privilege |
| Voting software | Unpatched flaws and bad configs | Paper trail and pre-election testing |
| Supply chain | Tainted parts before deployment | Inventory, isolation, vendor review |
| Polling sites | Threats that suppress turnout | Physical plans and rapid relocation |
| Insider access | One person with too many keys | Two-person control and observers |
What Counties Can Actually Use Before November 3
The useful core of the packet is the service menu. The agency is offering cybersecurity and physical security help at no cost. That includes assessments, hunting for known weaknesses, tabletop exercises, and advice on how to harden a network that was never designed to be a fortress. DHS, the plan notes, has kept backing those deliveries even when the political weather around the agency itself turned ugly.
If I were a clerk in a mid-size county, I would not start with a press conference. I would start with three unromantic tasks. First, know every device that will touch a ballot or a registration record, including the forgotten ones in a closet. Second, confirm that paper and pollbooks can survive a system crash without improvisation. Third, walk the physical plan for a hoax threat the way fire departments walk an evacuation. The day you need the binder is a bad day to discover the binder is from 2019.
- Inventory machines, pollbooks, and registration endpoints.
- Test paper backup and audit procedures with real staff, not just vendors.
- Lock down remote access and shared admin accounts.
- Rehearse physical threats, including site relocation.
- Document two-person custody from open to canvass.
Does that sound basic? Good. Basic is what holds when a rumor hits social media at 4 p.m. and a printer jams at 5.
Trust Is The Real Product, And It Is Fragile On Purpose
The stated goal is simple enough to print on a hallway poster. Americans should be able to trust voting systems and believe that physical safety measures will be in place when they show up at an assigned location. Notice the pairing. Cyber confidence and bodily safety are treated as one problem. That is overdue. A voter who feels watched or threatened does not care that the hash values on the tabulator look clean.
Trust, though, is not a setting you flip in a console. It is built from visible process. Can an observer see the count? Can a voter get a paper record? Can a county explain a discrepancy without vanishing into vendor-speak? Those questions are older than the current agency. They will outlive this plan.
I’ve found that the loudest arguments skip the middle of the process. People jump from “the system is perfect” to “the system is rigged” and leave no room for the ordinary mess of democracy: late supplies, confused poll workers, a scanner that hates a particular batch of paper, a registration record that was never updated after a move. Ordinary mess is not the same as coordinated theft. A serious plan has to hold both ideas at once.
A Short History Of Why This Keeps Coming Back
The agency itself was stood up in 2018, after years of warnings that election systems were critical infrastructure in name and leftover office tech in practice. Designating elections as critical infrastructure did not magically fund every county. It did create a federal partner that could knock on doors without waiting for a disaster. That partnership has always been awkward. Elections are local. Standards are uneven. Vendors are few. Politics is loud.
Every cycle since then has featured the same triangle: foreign probing, domestic suspicion, and local capacity that varies wildly from a well-staffed city to a rural office with two full-time people and a seasonal crew. The new plan does not dissolve that triangle. It tries to give the small office a menu that does not require a Fortune 500 security team.
There is a temptation to treat each new document as proof that the last one failed. Sometimes that is fair. Sometimes it is just how risk management works. You write the plan you can execute this fall, not the plan you wish you had written five years ago when the machines were purchased.
Software Failure Versus Software Manipulation
Those two problems get mashed together in public debate, and they should not. A software failure is a crash, a misread, a frozen screen, a batch that does not reconcile. Manipulation is someone using a flaw on purpose. Paper and audits help with both, but they help in different ways. Failure needs a restart path. Manipulation needs a trail that makes cheating expensive and visible.
Pre-election logic and accuracy testing is the unglamorous cousin of both. You run known ballots through the equipment and confirm the machine reports what you already know. If that step is skipped, rushed, or performed only by the vendor, you have already given away your best cheap defense. I would rather see a county publish its testing calendar than another slogan about resilience.
Resilience stack in plain language: Paper the voter can see Logs nobody can edit alone Two people on every sensitive step A tested plan when the site has to move
Physical Security Is Not A Side Quest
Cyber teams hate being told that a phone call can undo a week of patching. It can. A bomb threat, even a clumsy one, can close a site, reroute voters, and seed a narrative that the process is collapsing. The 2024 incidents are the reference point in the new text for a reason. Copycats do not need sophistication. They need a number that gets answered and a clerk who has no practiced reply.
Physical planning sounds like law-enforcement theater until you watch a precinct try to relocate on the fly. Ballot containers, pollbooks, accessibility equipment, chain-of-custody signatures, notification to voters already in line—those pieces have to move together. If they do not, you get two incomplete sites and a rumor that votes were lost in a van.
Training poll workers to treat a threat as a procedure instead of a panic is half the battle. The other half is telling the public, in advance, that relocation is possible and that a relocated vote still counts. Silence in that moment is how a nuisance becomes a legitimacy crisis.
What The Plan Does Not Settle
It does not settle who pays for replacements if foreign-made parts are deemed unacceptable. It does not settle how fast certification can move if a vendor needs to swap a component. It does not settle the argument about how much federal help local officials should accept. Those fights will continue after November, which is another way of saying the plan is a season of work, not a finale.
It also does not name the next specific foreign operation. Readers who wanted a villain of the week will leave hungry. I think that restraint is a feature. A public document that over-indexes on one country becomes outdated the first time a different actor gets loud. Categories age better than headlines.
One more gap is worth saying out loud. Communication with voters is treated as an outcome, not a craft. The best technical posture in the world still fails if a county cannot explain a delay in human language. Perhaps that is the next short document someone should write: a one-page script for when the scanner dies and the line gets restless.
A Practical Reading For Anyone Who Is Not An Election Official
If you are just a voter, the plan still has a translation. Look for paper. Look for observers. Look for a county that can describe its backup without sneering. Be wary of anyone who needs you to believe that one leaked screenshot proves a national steal, or that one official statement proves nothing can ever go wrong. Both postures are lazy.
Ask simple questions. Can this race be audited with paper? Who holds the keys to the registration system? What happens if this gymnasium has to close at noon? Those questions are not conspiratorial. They are how adults inspect a process that only works if strangers trust it for one long day.
I will be watching whether the no-cost services actually get used in the next few weeks. A plan that sits on a website is literature. A plan that shows up as an assessment in a county that asked for help is policy. The difference will not trend. It will show up in how boring Election Night feels, which, if we are honest, is the best possible result.
The Weeks Between The Memo And The Polls
There is not much calendar left. Early voting habits vary by state. Some places are already moving material. That compresses the window for anything ambitious. So the honest agenda is triage. Patch what can be patched. Staff the two-person teams. Print the extra paper. Walk the threat scenario once with the sheriff’s office and once without them, because the second walk always finds the missing radio.
Vendors will promise remote miracles. Some of those promises are real. Some are sales. A county that cannot explain a remote connection in one sentence should not flip that connection on in October. Isolation is not fashionable. It is how you keep a compromised laptop in the treasurer’s office from becoming a story about the vote itself.
And if a threat does hit a polling place, the measure of success is not a perfect quote on television. It is whether voters can still cast a ballot that can be checked later. That is a lower bar than the rhetoric. It is also the only bar that still means something after the cameras leave.
Closing Notes From A Skeptical Optimist
I do not believe a thirteen-page plan will end the argument about elections in America. Nothing that short could. I do believe a short plan that insists on paper, pairs, observers, and practiced physical response is pointing at the right pile of work. Fancy language about resilience is cheap. A signed seal number on a ballot box is not.
The midterms will be noisy. They always are. The test is whether the noise is about candidates and turnout, or about a process that cannot show its work. Officials now have a public checklist. Voters have a way to judge whether that checklist was treated as homework or as decoration. That is about as much clarity as a fall election usually gives anyone, and it is more than we had last month.
Show up. Watch the ordinary steps. Ask for the paper trail if your jurisdiction claims to have one. The rest of the drama will arrive on its own. It always does.
