
The right’s love affair with Alex Karp and Palantir was always a short-term fling that risked ending with a restraining order and a digital file on every GLOCK owner in America. When Palantir rolled out ImmigrationOS for ICE, plenty of people on the right treated it like a based Christmas present. Finally, someone with serious software was going to help enforce the border. Karp, the longtime Democrat who spent years talking progressive while cashing government checks, was suddenly the guy who “got it” on immigration. Stocks went up. Memes got posted. The surveillance architecture got thicker. For a few months it felt like a win. The same people who had spent years warning about Big Tech bias and government databases suddenly decided that this particular database was fine because the targets were the right ones. That is the trap - government power and products don't vanish just because the other side won some elections.
ImmigrationOS is not some minor tracking app. It is the Immigration Lifecycle Operating System—an ICE-funded platform, started with a roughly $30 million contract in 2025 and later expanded with additional maintenance and support funding, built on Palantir’s existing Gotham and Investigative Case Management systems. Its job is targeting and enforcement prioritization, self-deportation tracking in near real-time, and full lifecycle management from identification to removal. It pulls data from federal databases, commercial sources such as LexisNexis-style investigative products, and whatever else the agency can feed it, then surfaces leads with address confidence scores, maps, and dossiers so officers can prioritize and act. The ELITE component, for example, maps potential targets geospatially and ranks them. Agents get a picture of where people are likely to be, who is connected to whom, and which leads are worth the resource expenditure. The same general-purpose data fusion and analytics that lets ICE find people also lets any future administration find other people. Tools do not have politics. The people who task them do.
Palantir’s Gotham platform has been in use for law enforcement and intelligence for years. It turns disparate records into objects inside an ontology—people, vehicles, addresses, events—then links them so analysts can see relationships that would otherwise stay buried in separate systems. Entity resolution collapses duplicate records. Link analysis draws networks. Geospatial tools put everything on a map. Predictive scoring and prioritization layers sit on top. Once the ontology is built and the data feeds are connected, changing the mission is largely a matter of changing the queries and the policy guidance. The infrastructure does not care whether the list is “visa overstays” or “people who bought an AR last year.” That is the core capability.
Karp’s own history makes the pivot unsurprising. He has long described himself as a progressive and lifelong Democrat. He has bragged that Palantir’s software stopped “innumerable” terror attacks in Europe and that the only reason people are not goose-stepping between him and the audience is his product. In other words, he has framed his company’s work as essential to suppressing the political consequences of those attacks—consequences that, in his telling, would have strengthened far-right movements. The same platform that fuses data for one mission can be pointed at another. Open Source Defense noted the dual-use reality plainly: if the government ever decides to go hard against its own citizens, it will use the software it already bought and refined. As one observer put it in that discussion, “If the government ever puts us in camps, they’ll use Palantir’s software to do it.” Gun control is just another mission set. The data is already being collected and normalized. The ontology is already built. The right spent a couple of years treating that fact as a feature. It is a feature until the customer list changes.
The Misplaced Trust Was Obvious the Whole Time
The right’s willingness to cheer Palantir because it was useful against illegal immigration was the classic “enemy of my enemy” trap, with the usual suspects only focused on number go up. Yay capitalism and all but if you are selling out your nation and your heritage, stop. Palantir has always taken money from whoever writes the check. That is normal for a large government contractor - GLOCK sells to tin-pot dictators in New York State, for example. What is also normal is that founders prioritize customers and missions that align with their own politics and worldview. Karp has spent years signaling progressive priors even while defending controversial contracts. When the political winds shift, the company will shift with them. This is not a conspiracy. It is how large businesses run by people with strong views operate. If we here at RGG ran a company that sold high-end analytics and data mining takes, we'd cut better deals with customers who shared our 2A radicalism and charge the opposition full freight—or refuse the work entirely if the optics were bad enough. Anyone who has run a business understands this. Pretending otherwise is how you end up funding the tools that later get used on you.
Look at the company’s broader track record. Palantir has worked with defense and intelligence customers for two decades. It has also worked with ICE on case management since well before the current ImmigrationOS contract. The technology is dual-use by design. A system that can fuse border-crossing records, benefit applications, commercial identity data, and field tips for immigration enforcement can fuse NICS queries, state permit records, commercial purchase data, and ALPR hits for a different enforcement priority. The engineers who built the prioritization engine do not have to rewrite the codebase when the target list changes. They adjust the ontology mappings and the scoring rules. That is the product. A single pane of glass that today shows Rogelio Ocasio-Cortez de la Oca is a rapist from Guatemala could tomorrow show John Smith from Knoxville has a Daniel Defense AR and an opinion.
The stock price is not a moral compass. Watching shares climb while the underlying technology matures into a general-purpose targeting system is not the same as winning. Plenty of people on the right watched the valuation and decided that meant the company was on their side. Valuation measures how much money the government and other customers are willing to pay for the capability. It does not measure whose list the capability will be pointed at next year. The old line, commonly attributed to Lenin (though the precise origin is disputed and the exact wording varies), still fits:
The capitalists will sell us the rope with which we will hang them.
The American right has to get better at evaluating who it backs. This isn't a condemnation of capitalism, but an appeal to use your long-term capitalistic instincts for your long term benefit. Temporary utility is not the same as alignment. The next customer can rename the project and point the same machinery at a different list. Cheering the rope sale because you think you will be the one holding the other end is a bad bet when the seller has already told you, repeatedly, what his politics are. Karp isn't sneaky about it. When someone tells you who they are, believe them.
If That Trifecta Hits and GunOS Comes Online
Imagine 2028. The GOP loses big. A Democratic trifecta takes the White House, House, and Senate. The same people who spent years calling for more federal power over firearms now have the votes and the executive branch. They do not need to invent a new surveillance system from scratch. They have Palantir’s platforms, the Flock ALPR network we have already documented, state-level permit databases, commercial data brokers, and a growing pile of motivated private actors who treat gun ownership as a public health emergency. The result is a digital panopticon aimed at people who own guns. By the time the courts realize what's going on, there's target lists and boots on the ground. SCOTUS, even if it retains today's balance, ruling against the heavy hand of the executive branch is scant comfort since the damage is done.
It is reasonable to assume Palantir has already ingested or can readily ingest NICS data from the FBI, the ATF’s National Firearms Registration and Transfer Record for NFA items, state concealed-carry and purchase records where they exist, records from out-of-business FFLs stored in Martinsburg, and private commercial feeds. Flock cameras already log plates near gun shops and ranges. Credit-card and loyalty data, social media, and other open sources fill gaps. The same targeting logic built for ImmigrationOS—prioritization, confidence scoring, geospatial leads, lifecycle tracking—becomes GunOS. Analysts query for people who bought an AR, visited a range three times in a month, posted something about the Second Amendment, or simply own a GLOCK in a jurisdiction that just passed a new “machine gun” definition. The system does not need to be perfect. It needs to be good enough to generate lists and direct resources.
The surrounding ecosystem makes it worse. Groups like Everytown have already pressured AI companies to restrict firearms-related information, to flag conversations, and to refer cases to law enforcement. OpenAI has tightened rules around weapons content for minors and limited reconstructions of past attacks. Other firms face similar pressure. Anti-2A people, via their progressive podcasters and political mouthpieces, have already made it known they want to drag Elon down in lawfare, which could jeopardize Grok. In a trifecta environment those private systems become additional feeders and filters. A query that starts in a government ontology can pull commercial location data, ad-tech identifiers, and platform reports. The panopticon is not one product. It is the stack.
Picture an ordinary guy or gal who posts on X or local forums about training with their rifle and supporting the Second Amendment. We're not talking about a guntuber or so-called "milita" leader. The person is just a homeowner who bought a suppressor after the tax went to zero and talks about it. Maybe they are even a GOA member who just transferred the can out on a 4473. Under a hostile administration, a fused "GunOS" query flags this person. Flock data shows repeated trips to a gun shop. Illegally-retained NICS data and commercial records confirm recent purchases. Social media fills in the political views. A local fusion center or federal task force gets the package. Agents show up at 5 a.m. with a warrant built on the fused data. The raid is justified as “public safety” or “compliance check.” A good person not expecting the cops at 5 AM may respond to the noise with a gun in hand. You know the rest. Neighbors film it. The clip circulates. The point is the example, not the conviction rate. "If you are this person, the government will try to kill you and half the population will cheer." Everyone else who owns the same configuration now has a data point about what happens when the system notices you.
Or skip the outspoken part. Take a normie who owns a GLOCK and an AR because he lives in a rural area and has the space and time to enjoy some serious shooting. He has never posted about politics. He just owns common guns. In the same trifecta environment, the system is tasked with identifying high-volume owners or people who bought “banned” configurations before the hopefully-hypothetical cutoff. A list is generated from purchase records, range visits logged by cameras, and any state database that still exists. Local police, under federal pressure or simply following new state rules, conduct “welfare checks” or enforcement actions. The goal is deterrence through spectacle. Again, half the population will cheer when the normie is burned out of his home. The data fusion makes the spectacle efficient. Palantir and the surrounding stack turn scattered records into actionable targets. That is the point of the architecture. You do not need a manifesto on your wall. All the system wants is "gun owned by this person of this demographic".
How Second Amendment Radicals Fight This Without Becoming Luddites
This does not require smashing computers or going full Sarah Connor. It requires treating data the way competent people already treat physical security: assume hostile collection and reduce the attack surface. Use cash for gun and ammo purchases when practical. Use Bitcoin for the rest, self-custodied, because it does not ask a risk committee whether the Second Amendment is still fashionable this quarter. Vary routes. Support shops and ranges that understand the threat. These are basic citizen OPSEC steps, not conspiracy theories. The same discipline that keeps a range bag from becoming a tracking beacon also keeps a purchase from becoming an easy node in someone else’s target database.
Support companies whose leadership and incentives are less likely to flip the switch the moment the party in power changes. SpaceX and Starlink provide resilient communications that do not depend on a single carrier’s goodwill. Thus far, Elon has demonstrated a certain tolerance of the 2A community, and in our opinion is that it's good as one can expect from the richest man on earth. Anduril, founded by people who have been more open about defense and individual rights, is another example of tech built outside the traditional progressive consensus. Palmer Luckey explicitly did what he did partially to get back at Facebook. Look for other builders who treat free people as the default rather than a problem to be managed. The goal is not to create a new monopoly. The goal is to keep options alive so that critical infrastructure does not all route through firms that treat your existence as a compliance problem.
Reach out to elected officials and explain, in plain language, that a tool built for one “threat” can be re-tasked for another. Immigration enforcement architecture becomes gun enforcement architecture with a change in priorities. That is not abstract. It is how software works. Ask them what happens to the data contracts, the sharing agreements, and the analytic tools when the political leadership changes. Most of them have not thought past the current news cycle. Make them think past it.
Build local resilience. Second Amendment Sanctuaries and similar local policies raise the political and practical cost of federal or state overreach. Talk to county sheriffs, commissioners, and state legislators. The more the enforcement network is fragmented and dependent on local cooperation, the harder pure top-down targeting becomes. A national system still needs local bodies to execute. Raising that cost is useful even if it is not absolute.
You also have a right to compute. Montana’s Right to Compute Act and similar efforts treat ownership and use of computational resources as an extension of property and free expression rights. Jensen Huang didn't get sent off from the company he founded because he was unfaithful to his wife - heck, each and every member of the NVIDIA board is likely a philanderer - Jensen got sent off because he supported open source models and local AI. Jensen would gladly sell you a DGX box, and wish you well on your quest.
Democrats have already signaled the opposite direction: a flood of bills for independent AI testing, liability regimes, data-center energy rules, and even proposals to pause or ban advanced systems ownership - except from the government of course. Some of those measures will be framed as safety. Others will function as barriers that keep serious capability inside government-approved channels. Oppose any framework that reserves advanced tools for the state while restricting private citizens. The same code and hardware that can run a small business or a 3D-printed firearm design can also be used by free people to stay informed and organized. Restricting that to government entities only is not safety. It is control. We already have articles on this site about the connection between the right to make your own firearms and access to computing tools. That connection does not get weaker when the surveillance stack gets stronger.
An Armed Citizen Can Still Say No
The situation is serious. The technology exists. The political incentives exist. The data is being collected and normalized right now. That does not mean the outcome is fixed. An armed citizen can say no and back it up if needed. The possession of arms is the practical check on government that decides it no longer needs consent.
As one classic formulation puts it: an armed man is a citizen; an unarmed man is a subject. Joseph Story noted that the right to keep and bear arms is a strong moral check against usurpation. The power is not theoretical. It is the reason free societies have historically been harder to fully dominate than disarmed ones. Local organization, sound money, resilient communications, and the willingness to treat surveillance architecture as hostile by default all raise the cost of the panopticon. The people who already understand this—the ones documenting the cameras, the NFA fights, the data brokers—are still here. The tools cut both ways. Use them on your terms, support the builders who will not sell the rope, and remember that an armed populace that refuses to consent is still the hardest problem any central authority has ever tried to solve.
The midterms showed that electoral setbacks happen. They also showed that the underlying knowledge, manufacturing capacity, and cultural shift do not vanish because a cycle went the other way. The same networks that have been tracking NFA cases, ATF overreach, and surveillance expansions are still operational. The response is not despair. The response is tighter OPSEC, better local ties, clearer communication with the people who write the laws, and continued support for the technical capacity that keeps free people able to say no. Trust the people who will not flip when the contracts change. Keep the hardware and the knowledge local. And never forget that the final veto does not live in a database.
The practical reality of these systems is that they reward volume. ImmigrationOS and the older ICM/Gotham stack were built to process large numbers of records and surface a manageable number of high-priority leads. That same math works for any population the government decides is a priority. Once the feeds are connected—NICS transaction logs, NFA registration data, state license records, commercial data-broker files, Flock ALPR hits, and whatever private platforms are willing to share under a new legal or political regime—the system can generate ranked lists on demand. Confidence scores on addresses, link analysis showing family members or frequent contacts, and geospatial clustering that identifies “target-rich” areas all transfer directly. The analysts do not need a new degree. They need a new set of policy rules and access permissions. History shows those are easier to change than the underlying software.
Karp’s European comments are worth sitting with longer than a single sentence. When he says the only reason people are not goose-stepping is his product, he is hyperbolically claiming that Palantir’s data tools prevented terror attacks that would have produced a political reaction strong enough to empower what he thinks is the far right. That is a statement about political outcomes as much as it is a statement about security. It treats the suppression of a certain kind of political movement as a benefit of the software. The same logic applies in reverse. A future administration that views gun owners, or a subset of them, as a political problem rather than a constituency with rights will find the same software useful for a different purpose. The technical capability does not change. The definition of the problem does.
Don't Simp For Tech Bros
People on the right are not uniquely bad at this. Plenty of institutions have convinced themselves that a vendor who is useful today will remain useful tomorrow. The difference is that the right has spent years correctly identifying the risk of permanent data systems and then, for a brief period, set that knowledge aside because the current targets looked good. ImmigrationOS was real work. It produced real leads. It also produced a more mature data environment that future users will inherit. The contract language, the sole-source justifications, and the existing integrations with other federal systems all make the handoff easier. When the next administration wants a different priority list, the infrastructure is already there.
Founders and executives have preferences. Karp has been clear for years that he sees himself as progressive even when he defends contracts that progressive activists dislike. That combination is not unusual in government contracting. Companies take the work, defend it on competence or national-interest grounds, and still maintain the political identity of their leadership. When the political identity and the customer base realign, the product follows. The rope quote is useful precisely because it is blunt. The seller does not have to hate the buyer. The seller only has to be willing to sell to whoever is in a position to pay and to task the system. Once the system is sold and integrated, the original buyer does not control the future tasking.
Stock performance creates its own distortion. A rising share price is visible and easy to cheer. The downstream risk is diffuse and easy to ignore until the tasking changes. Plenty of observers treated Palantir’s growth under the current administration as evidence that the company had switched sides. The more accurate reading is that the company sold a capability that multiple administrations can use. The valuation reflects the size of the market for that capability, not the political durability of any particular customer relationship. Treating the stock chart as a loyalty test is how you end up surprised when the next customer writes a larger check.
The 2028 scenario does not require a sudden technological breakthrough. It requires a change in priorities and access. A Democratic trifecta would inherit the existing Palantir deployments, the Flock network, the commercial data contracts, and the legal authorities that already allow significant data sharing between federal, state, and local agencies. Adding or expanding access to NICS, NFA, and state firearm records is a policy and legal question more than a technical one. Once those records are inside the same ontology that already contains immigration, criminal, commercial, and location data, the queries write themselves. “Show me people who bought an AR in the last two years and whose plate was logged near a range more than four times” is not science fiction. It is a report.
Private actors accelerate the process. Everytown and similar organizations have already pushed AI companies to limit firearms content, to monitor conversations, and to refer certain cases to law enforcement. Those policies create additional data exhaust. A user who asks a chatbot about a particular firearm configuration, or who is flagged by a platform for “weapons-related” discussion, generates a record that can later be requested or purchased. In a hostile political environment those records become another feed. The government does not have to build every sensor. It can buy or compel access to the sensors that already exist.
The outspoken scenario is the one people picture first because it is cinematic. The quiet scenario is the one that scales. A system that can rank every owner of a particular configuration by purchase date, range visits, and address confidence does not need the target to have posted anything. It needs the target to have bought the gun through a channel that left a record and to live in a place the cameras already cover. Most people who own common rifles and pistols fit that description. The spectacle of a few high-profile raids does the political work. The bulk of the list does the administrative work. Both are products of the same fused data environment.
Note, this is far from a condemnation of capitalism. As a matter of fact, it's an ardent support of such. Second Amendment Radicals have the ability and discretion to spend with those who support them, and build their own thing. Being armed means we can say "no", back it up with meaning, and go to someone else. Palantir can kick rocks.
note: I wrote this with the aid of Grok. Let me know what you think. With Grok I can output faster, but it may not seem like "me" as much. Without Grok, you get "me" but slower.

Get Yourself A Gun And Learn It
This is a firearms blog, so we're going to recommend guns. Of course don't forget other things like medical, comms, doing your Area Study, and so on. There's better sources than RGG on those topics, but for firepower, we have you covered.
Ideally you'll at least want either a basic rifle, i.e. an AR-15, or a pistol, i.e. a GLOCK 19, to start. Both are exceedingly common firearms with plenty of accessories, spare parts, training, and support. However, in some restricted areas, both an AR and a GLOCK are hard to get easily, and you may have to settle for a shotgun for quickly arming yourself. It beats a sharp stick. Note, the in-house counsel wants to remind you that we're talking the legal acquisition of guns here. We're still working on abolishing gun control, and we don't want you in jail. Unfortunately you'll have to play ball in this regard. But anyways, here are some basic suggestions for getting armed.
| Firearm | Cost | Buy Now |
|---|---|---|
| Radical Firearms RF00028 AR-15 | $469 | Buy Now! |
| GLOCK 19 Gen6 9mm Pistol | $620 | Buy Now! |
| Mossberg 590 Shotgun | $621 | Buy Now! |
Learn how to make guns, as well.
| 3D Printer | Cost | Buy Now |
|---|---|---|
| Ender Creality 3 V3 SE | $186 | Buy Now! |
Support This Site
Operating Regular Guy Guns and bringing you quality content costs money, money that I am more than happy to spend, even after all these years! I get the occasional sponsor, but bills still have to be paid, ha ha.
You'll see the articles peppered with affiliate links. I get a few pennies when you make your purchases via my links.
-
Buy more guns at Palmetto State Armory
-
Buy more guns at Bear Creek Arsenal
-
Buy more guns at Brownells

TODO
Broken links in old articles
Addendums to old stuff
Buy more guns and make more guns
Podcast? AI killed the podcast.
Donate Bitcoin: 36wKfH7wgQQna6BByvAe8oiEmdqREUXuYQ