Peter Thiel’s Founders Fund backs a controversial company called Persona Identities. Persona provides “age verification” to consumer platforms including Discord, OpenAI, Roblox, LinkedIn, and Lime.

Researchers have exposed Persona’s entire frontend architecture, with nearly 2,500 files at 53MB, sitting open on a FedRAMP government endpoint tagged with codenames from active intelligence programs. Fortune has the details, and the researchers published their findings on Twitter.

Yeah. FedRAMP. Thiel exposed, again.

The files show Thiel’s Persona version of surveillance runs 269 distinct verification checks on users. These include expected facial recognition against watchlists, but also screening against lists of politically exposed persons, and adverse media categorization across 14 categories including terrorism and espionage. The system assigns risk and similarity scores to every person.

EVERY person.

He has deployed a biometric surveillance pipeline, connected to Palantir money and dubious motives, through consumer products It’s a thin front to government intelligence infrastructure.

And he has it running right now on platforms used by hundreds of millions of people, to manipulate children by tricking them into believing they are just proving their age, unaware it’s literally Big Brother.

Discord Deployed, Not a Test

Earlier this month Discord defaulted all accounts to teen-safety settings.

ALL on by DEFAULT.

The announcement forced users into a binary: verify age using the Persona gate or lose platform access.

Discord hosts roughly 200 million monthly active users. Its population includes political organizers, activist communities, open source developers, security researchers, journalists and their sources, student groups, and gaming communities that overlap heavily with military-age males.

Thiel was targeting precisely the populations his private intelligence services would want to enumerate and categorize, as a Palantir-adjacent system.

Discord users specifically chose to join as a pseudonymous platform. They use handles, avatars, anonymous identities. Persona maps real biometric identity onto those pseudonymous accounts.

Before Thiel slammed in Persona, a Discord user was a handle. After the Thiel Persona gate was in place, that handle is linked to a face, a government ID, a watchlist status, a politically-exposed-person score, an adverse media categorization across terrorism and espionage categories.

The pseudonymity that made Discord attractive to its users is exactly what made it valuable to Thiel’s extremist ACTS 17 congregations as a collection target. The operation was to de-anonymize the precise population that specifically chose anonymity as protection from Thiel.

The Constitutional Bypass

The Fourth Amendment constrains what the government can compel from citizens. The political and legal cost of requiring 300 million people to submit biometrics to a federal database would be astronomical.

Thiel routed around American legal rights entirely. A private company provides “age verification” to platforms people already use. Users submit biometrics to Discord, not directly to the government. Persona processes the data as a private service provider.

The fact that data landed on FedRAMP government endpoints tagged with intelligence program codenames gives away the corruption. At no point does the state compel anything from anyone, because the surveillance becomes the product. The Third Party Doctrine that information voluntarily given to a third party loses Fourth Amendment protection does the rest.

This is obviously a laundering mechanism to undermine constitutional constraints on state power. The private sector collects what the state cannot legally demand. Commercial authorization frameworks move the data to government infrastructure. The child safety language is disinformation to make opposition politically suicidal.

The Franchise Model

The Thiel portfolio has to be seen as a vertically integrated surveillance supply chain to understand his documented upbringing as a Nazi.

Compulsion	Thiel backs politicians — Vance, Trump — who support age verification mandates. Bipartisan cover is guaranteed because child safety is politically inarguable. These mandates create legal pressure for civilians to submit biometrics through consumer products they already use.
Collection	Persona embeds in platforms where billions of users live. Age verification framing makes biometric submission feel routine and protective. Users submit to Discord or Roblox, not to the government. Every parent who consents on behalf of a minor is enrolling a child in a system whose back end they cannot see.
Processing	Persona’s 269 checks are a civilian threat assessment engine. Watchlist matching, politically exposed persons screening, adverse media categorization, risk scoring. This is intelligence-grade analysis running on people who think they are confirming a birthday.
Bridging	FedRAMP authorization formalizes the government channel. Persona CEO Rick Song frames this as “workforce security.” The infrastructure is bidirectional. Once civilian biometrics sit on government-authorized endpoints tagged with intelligence program codenames, the distinction between “workforce verification” and “population surveillance” is a configuration setting.
Action	Palantir correlates identity data with intelligence streams. Anduril operationalizes it for defense and border enforcement. The supply chain runs from a teenager’s selfie on Discord to an actionable targeting package.

Each platform is an independent collection node. Each has its own justification. Each feeds the same backend. Each can be severed independently if exposed without disrupting the others.

Discord got burned and got cut off.

Persona continues operating through OpenAI, Roblox, LinkedIn, and Lime.

The Rollback Story

Last year, hackers accessed 70,000 government IDs collected through Discord’s previous verification vendor, 5CA. Discord’s response to that breach was to switch vendors — to Persona, which has deeper capability and a direct pipeline to government infrastructure. The breach provided cover for upgrading the collection system.

When researchers published that Persona’s architecture was sitting on government endpoints, the cleanup began immediately. Discord cuts ties. Both companies coordinate messaging to say the partnership lasted “less than a month.” Song tells Fortune the exposed files are just “uncompressed frontend” and that internally this was not considered a major vulnerability. He posted screenshots of an email exchange with the researcher, claiming the implication of connections between Persona, Palantir, and ICE has led to threats against company employees.

The data that flowed during that month already flowed. The biometrics already hit the government endpoint. The watchlist checks already ran. The 269 verification checks already executed against every user who submitted. Dissolving the partnership does not un-run those checks.

A since-deleted version of Discord’s FAQ on age verification contradicts the company’s claims about data retention, stating information would be “temporarily stored for up to 7 days.”

Discord amended its universal age-verification announcement to say verification would be “optional” — unless users want access to age-restricted servers and channels, which means most of the platform’s actual functionality. This is compulsion through architecture rather than law.

A CEO Knows

Song was attacked for lacking a profile photo on his own LinkedIn page — the same LinkedIn whose identity verification Persona handles. His response: “It’s dystopian that we want people to facedox themselves to everyone to be real online.”

The CEO of a facial recognition verification company argues against making faces publicly visible — while his product links faces to identity databases, intelligence watchlists, and government endpoints for hundreds of millions of people. He knows linking faces to identity databases is dangerous. He built a system that does it at population scale. His defense is that he personally should not have to participate.

That is hierarchy, stated plainly. The system makes identity transparent downward and opaque upward.

Recognition Time

In 2009, Peter Thiel wrote that freedom and democracy are incompatible. He was describing a design constraint. Democracy creates legal barriers to population-scale surveillance. Market mechanisms achieve what democratic governance prohibits.

Age verification mandates are spreading across state legislatures. Persona is embedded in platforms used by hundreds of millions of people. FedRAMP authorization formalizes the government pipeline. The administration dismantling institutional checks — inspectors general fired, DOGE gutting federal agencies, loyalty tests replacing competence — is the same administration receiving the output of this biometric collection network.

The 269 checks exist now. The watchlist screening exists now. The politically exposed persons matching exists now. The adverse media categorization across terrorism and espionage categories exists now. The intelligence program codenames exist for a reason that Discord isn’t disclosing. The question is whether anyone treats this Thiel story as the emergency it already is.

While Tesla was busy confessing to the California DMV that Autopilot was never real, FuelArc ran the numbers on the Cybertruck’s fire fatality rate against the Ford Pinto — the historic benchmark for corporate greed killing customers.

The Cybertruck is 17 times worse.

SEVENTEEN TIMES WORSE THAN A PINTO

Five fire fatalities in 34,438 vehicles gives the Cybertruck a fatality rate of 14.52 per 100,000 units.

The Ford Pinto, across a decade of production and 3.17 million vehicles, managed 0.85.

The Pinto became a national scandal.

The Cybertruck army of Elon Musk adherents sent a death threat to the journalist who did math.

The NHTSA still hasn’t crash-tested the Cybertruck. Tesla still hasn’t released official delivery numbers. And the company that just admitted its entire “driverless” branding was a lie is simultaneously selling an untested vehicle that burns its occupants at a rate that would have gotten any other manufacturer shut-down and hauled before Congress.

The Pinto at least had the excuse of being a Ford, and cheap.

The Cybertruck, marketed as a “survivalist” design, costs six figures for a barely operational bucket of half-baked bolts that kills faster than the Pinto.

Tesla accepted every correction, then sued to erase the record that forced them.

Tesla officially admitted it never had Autopilot. Not on Twitter. Not in SEC filings. In the most boring and quiet way possible: compliance with a California DMV order to stop lying to the public.

On February 17, 2026, the California DMV announced that Tesla had removed “Autopilot” from its marketing and discontinued it as a standalone product in the US and Canada.

This followed a December 2025 ruling by an administrative law judge (ALJ) who found that Tesla’s use of the term was misleading and violated state law. The ALJ called Full Self-Driving a name that is “actually, unambiguously false.”

And water is wet.

The DMV’s own language is unsparing when you see how they officially describe vehicles equipped with those ADAS features:

…could not at the time of those advertisements, and cannot now, operate as autonomous vehicles.

Elon Musk lied. He never, ever corrected the lies, even as hundreds were killed.

Tesla had every opportunity to fight.

They could have contested the findings, challenged the ALJ’s authority, appealed through administrative channels. Instead they quietly complied within the 60-day window, stripped the branding, and restructured their entire ADAS product line around the order. All on the same date.

Their compliance is a HUGE confession.

The Lawsuit That Explains Everything

Four days before the DMV announced Tesla’s compliance, Tesla filed suit in Los Angeles Superior Court to erase the ruling entirely. The company alleges the DMV “wrongfully and baselessly” labeled it a false advertiser.

Tesla’s legal argument is that no California customer was ever actually confused about whether their car could drive itself. This is a company that accepted the marketing was indefensible, made every change the state demanded, then sued to eliminate the official record that forced those changes.

They don’t want the word “Autopilot” anymore, because the ruse is up. They want the finding gone. They want their victims to be at fault.

The reason is straightforward.

Tesla has told investors it has 1.1 million “FSD subscribers” and its entire valuation thesis depends on becoming a robotaxi company. A formal, on-the-record finding of false advertising about self-driving capabilities is a powerful staff of justice handed to every plaintiff’s attorney in the country. The lawsuit is all about liability containment.

The Damage Already Done

The liability is real. A federal judge just upheld a $243 million verdict against Tesla in a fatal Autopilot crash — the first major plaintiff victory in an Autopilot wrongful death case. Tesla had rejected a $60 million settlement before trial.

Since that August 2025 verdict, the company has quietly settled at least four additional crash lawsuits rather than let more juries hear the evidence.

In January 2026, Tesla was sued over a Model X crash that killed an entire family of four when the vehicle allegedly veered into oncoming traffic. A separate class-action involves customers who bought Full Self-Driving expecting their cars to become robotaxis and now want refunds. NHTSA launched a probe in October 2025 into 2.88 million Tesla vehicles after connecting 58 incidents to FSD, including vehicles running red lights and driving into oncoming lanes.

Dozens more cases are working through the courts, because Tesla is a death trap based on lies.

The Defense That Convicts

Tesla’s arguments in the lawsuit deserve attention for what they reveal.

First:

It is impossible to buy a Tesla equipped with either Autopilot or Full Self-Driving Capability without seeing clear and repeated statements that they do not make the vehicle autonomous.

This is an admission that the product names were always misleading enough to require disclaimers, which existed precisely because the branding said something the technology couldn’t deliver.

I’ve written before how this directly contributed to mass death. No marketing would have been far better than the fraud of claiming a capability, and then warning against it. It’s like walking into a room and saying “I’m not going to accuse this person of murder” and then pretending you never said anything people should think.

Second: the DMV had known about the “Autopilot” branding since 2014 and “Full Self-Driving” since 2016, and therefore shouldn’t be able to act now. This is the argument that you’ve been getting away with a crime for so long that your crimes should be legalized. Tesla tried this same defense in 2023. It didn’t work then either.

The ALJ addressed the consumer confusion argument directly in her ruling. The DMV’s authority to regulate vehicle advertising:

…does not depend on evidence that any particular advertising actually has deceived or harmed any person.

The state can act to prevent deceptive advertising. It doesn’t need to wait for a body count, though Tesla has provided one anyway.

Supervised Self-Driving

Tesla’s retreat position is “Full Self-Driving (Supervised)” as an obnoxious oxymoron that exists because they needed language that preserved the FSD brand while denying FSD has any brand value. It’s like selling a life preserver (concrete shoes) and then shrugging at all the drowned victims. Supervised self-driving is driving with a huge marketing budget and propaganda office working round the clock. The parenthetical does the liability shift while the product name does the toxic selling.

Meanwhile, Tesla is building the fraud of a Cybercab (1950s science fiction of no steering wheel and no pedals) on the premise that full autonomy is imminent. The company that just admitted it can’t call its current system “Autopilot” because that overstates its capabilities is simultaneously manufacturing a car that pretends to have no manual controls at all.

This is a contradiction Tesla is betting everyone and especially investors won’t care about.

What Compliance Means

Tesla complied with the DMV order, removed the branding, restructured its product line, and then sued to pretend none of it happened. The sequence tells you everything. The marketing was false. The technology was oversold. People died in cars they were told could drive themselves. The state called it what it was. Tesla accepted the correction and is now trying to shred the receipt.

The ALJ noted in her decision that without the threat of suspension, Tesla offered:

…no reason for the DMV to expect that respondent will alter the Autopilot name, or will act to avoid continuing its misrepresentations to the public.

She is right.

Tesla didn’t stop because it agreed. It stopped because California is its largest EV market and a 30-day sales ban was unacceptable. Compliance was a business decision. The lawsuit is the tell that it was never a moral one.

Autopilot is gone because a judge ruled it was always a lie and Tesla chose market access over defending the claim.

That’s the core admission. Tesla is and has always been fraud. Elon Musk is worthless.

Everything else is lawyering.