The Court That Said Stop

Not to review its procedures. Not to seek guidance. Stop. On August 17, 2006, Judge Anna Diggs Taylor of the Eastern District of Michigan held that the warrantless wiretapping program the President had authorized in October 2001 violated the First Amendment, the Fourth Amendment, the separation of powers, and the Foreign Intelligence Surveillance Act. She issued a permanent injunction. She wrote the line that got printed everywhere: "There are no hereditary Kings in America and no powers not created by the Constitution. So all 'inherent powers' must derive from that Constitution."

She was correct. It did not matter.

The government appealed. Within weeks the injunction was stayed pending appeal and collection continued uninterrupted. On July 6, 2007, a 2–1 panel of the Sixth Circuit vacated her ruling without touching the merits — the plaintiffs, it held, could not establish standing, because they could not show they personally had been surveilled. The dissenting judge would have reached the merits and affirmed on the FISA violation. In February 2008 the Supreme Court declined to hear it. By then Congress had already legalized the program.

That is the sequence. It is worth walking slowly, because almost nobody involved did anything wrong, and the outcome was total.

i · the doorway in the load-bearing wall

Start with what made Taylor's ruling possible at all, because it was narrower than the quotable sentence suggests.

The government's primary defense was the state secrets privilege: the case cannot be tried, because trying it would require disclosing whether the program exists and whom it touched. Taylor rejected that as to the wiretapping claim for one reason only — the President had already publicly confirmed the program's existence after the New York Times revealed it on December 16, 2005. She accepted the privilege as to the companion data-mining claim, and dismissed it, because that part had not been conceded.

Read that again as mechanism rather than as law. Judicial review of the program was available in exactly the region the executive had already surrendered on television, and unavailable everywhere else. The check did not fail. It functioned precisely to the edge of the disclosure and then stopped.

Standing did the rest. To sue, you must show injury. To show injury from secret surveillance, you must know you were surveilled. Whether you were surveilled is the secret. The party holding the secret is the party being sued.

This is not corruption. It is a closed loop, and it is load-bearing. Every judge in the sequence applied doctrine as written. The Sixth Circuit's standing analysis was defensible; Taylor's own opinion was criticized as thin even by scholars who agreed with her conclusion, which handed the panel a comfortable off-ramp. Nobody had to be captured. The architecture did the work.

One qualification, because a structural frame slides into absolution if it isn't held. The four years between October 2001 and December 2005 were not weather. Routing around FISA by presidential authorization, and keeping that secret, was a decision made by identifiable people, with lawyers who wrote memos to make it sayable. That part was a choice. What the architecture explains is not why the choice was made but why it was never priced — why no plaintiff could form, why no remedy attached, and why the eventual institutional response was to ratify it. The structure did not commit the act. It absorbed it.

The Supreme Court confirmed the loop's durability in Clapper v. Amnesty International USA (2013), holding 5–4 that lawyers and journalists challenging surveillance under Section 702 had no standing because their fear of being monitored was "speculative" and their precautions were self-inflicted costs. Months later, the Snowden disclosures established that the collection was considerably broader than the plaintiffs had alleged. The Court had ruled the injury too speculative to litigate while the injury was occurring at industrial scale. The reasoning was internally sound. That is the point. A system can be perfectly coherent — every organ functioning, every doctrine consistent — and still produce a result that would be impossible to defend if the affected parties were in the room. They were not in the room, because they could not be told they were affected.

Compare a case where a court said stop and the machine actually stopped. In Youngstown Sheet & Tube v. Sawyer (1952), the Supreme Court held that Truman's seizure of the steel mills exceeded his authority, and Truman complied within hours. The usual reading is judicial courage. The structural reading is simpler: the steel companies knew they had been seized. The injury had an address and a plaintiff and a date.

Then set Ex parte Merryman (1861) beside it, because it breaks the rule and the break is the useful part. Merryman was maximally visible — a name, a date, a cell in Fort McHenry. Chief Justice Taney ruled Lincoln's suspension of habeas corpus unlawful. Lincoln ignored him. Merryman stayed put. So visibility is not sufficient. It is necessary. Seeing the taking is what makes a remedy formable at all; whether the remedy binds depends on whether defiance costs the executive something politically, and in the spring of 1861 it cost Lincoln nothing.

Which makes the surveillance case worse than either. Merryman could not get relief, but he could be a scandal — there was a constituency available, because anyone could see the cell. Warrantless collection defeats both conditions in one stroke: no plaintiff can form, because no one can learn they were surveilled, and no constituency can form for exactly the same reason. Youngstown is compliance. Merryman is defiance. What begins in 2006 is neither, and it is the more durable option.

Judicial review is a mechanism that requires a taking someone can see. Everything downstream — standing, ripeness, state secrets — is that requirement expressed in doctrine. Surveillance is the taking you cannot see. It is not that the courts declined to check it; the instrument does not fit the object.

ii · the previous excavation

Nothing above is a first. Taylor was enforcing the remedy from the last time this happened.

In 1975 and 1976, the Church Committee documented Operation SHAMROCK, under which the NSA obtained copies of international telegrams from Western Union, RCA Global, and ITT World Communications for roughly thirty years, and Operation MINARET, which watchlisted some 1,650 American citizens including Martin Luther King Jr., Muhammad Ali, and two sitting senators. The scandal was enormous. The response was structural and serious: Congress passed the Foreign Intelligence Surveillance Act on October 25, 1978, requiring a warrant from a dedicated court for surveillance of Americans, with criminal penalties for doing it otherwise.

That statute held for twenty-three years. In October 2001 it was simply routed around by presidential authorization, and the routing was secret for four years. The 2006 ruling was a court attempting to enforce the fix from the 1975 cycle, and being told it lacked jurisdiction to notice.

So the stratigraphy runs: abuse, exposure, outrage, statute, circumvention, exposure, ruling, vacatur, re-authorization. Roughly a generation per rotation. Each layer's reform becomes the next layer's inconvenience.

iii · re-authorization is the response

Here is the part of the pattern most coverage still misses. When a check fires on a security program, the executive branch does not typically comply, and does not typically defy. It launders.

Eleven months after Taylor's injunction, Congress passed the Protect America Act of August 2007. Eleven months after that, the FISA Amendments Act of July 2008 created Section 702 — programmatic, certification-based collection of foreign targets' communications from American providers, with Americans' communications swept in incidentally and retained — and granted retroactive immunity to the telecommunications companies that had participated in the program Taylor had declared illegal.

Not one wire was moved. The conduct was reclassified.

That is the most efficient response available to a system under judicial pressure, and it is why "unconstitutional" is a weaker word than it sounds. A ruling addresses a program's legal authorization. Authorization is the cheapest component to replace. The infrastructure, the contracts, the collection points, and the institutional appetite all persist through the swap, and afterward the program is not merely legal but democratically ratified, which is a stronger position than it occupied before the lawsuit.

The subsequent rulings follow the template. In ACLU v. Clapper (2015) the Second Circuit held that Section 215 had never authorized bulk telephone metadata collection — a clean plaintiff, a real merits ruling, twelve years into the program, arriving only because Snowden had supplied the standing the doctrine required. Congress responded with the USA FREEDOM Act, which relocated the records to the carriers and continued the queries. In United States v. Moalin (2020) the Ninth Circuit found the metadata collection likely unconstitutional and affirmed the conviction anyway. Every ruling lands correctly on a program that has already changed address.

Section 702 has been reauthorized in 2012, in 2018, and again on April 20, 2024 by the Reforming Intelligence and Securing America Act, which set a two-year sunset — April 20, 2026. Through the spring and summer of 2026 the authority has run on short extensions and failed votes while both parties negotiate reforms neither will finalize. Reform is now the permanent condition rather than an event.

And the design has advanced past the need for any of it. The Brennan Center's 2026 resource page on Section 702 contains the fact that closes the twenty-year arc more cleanly than anything a judge has written: collection will continue until March 2027 even if the statute lapses, because the certifications already issued by the FISA Court run for a year from issuance.

Be precise about what that is, because the tempting version is wrong. The program is not transcending law. It is running on a continuity provision working exactly as drafted, so that collection does not drop the instant Congress misses a deadline. Which is the more damning reading, not the lesser one. A sunset is not a stop. It is a twelve-month taper, designed as a taper in advance, by the same body that would have to be the one saying no.

That is what twenty years of institutional adaptation buys: no single institution's "no" — not a judge's injunction, not an appellate mandate, not the expiration of the statute itself — halts collection inside the year it is uttered.

iv · what cannot form a plaintiff

The government's case deserves stating, because the people who disagree with it rarely state it. Section 702 targets non-U.S. persons reasonably believed to be located abroad. Americans are not targeted. Their communications enter the collection incidentally, when they correspond with a target, and the live dispute is about what may then be done with what was swept in — retention periods, and queries run against the database using American identifiers. On its own terms this is a narrower program than its critics usually describe, and its defenders are not lying about it.

The reply is not that the description is false. It is that "incidental" is carrying enormous load for a category that includes everyone who has ever corresponded across a border, and no public number exists for how large that category is. The standing loop is why. Producing the number requires identifying the people; identifying them creates plaintiffs; plaintiffs are the thing the architecture is best at preventing. A program whose scope cannot be measured because measuring it would generate standing is not narrow. It is unfalsifiable, which is a different property and a worse one.

Which returns us to the useful question, and it is not whether the checks exist. They exist. They fired. Taylor was right on the merits, the Second Circuit was right on Section 215, the Ninth Circuit was right about the Fourth Amendment. Being right is not a mechanism.

The question is whether the affected can find out they were affected, because every other remedy is downstream of that one. Standing requires knowledge. Political accountability requires knowledge. Consent requires knowledge. A structure that reaches perfect internal coherence by keeping the injured party unaware of the injury has not solved the legitimacy problem; it has deleted the participants who would raise it and called the resulting quiet agreement. Legitimacy comes from including the affected, not from producing consistent doctrine about them in their absence.

But the twenty-year frame only earns its dateline if we say what changed. In 2006 the not-knowing was manufactured by secrecy: a covert program, classified, exposed by two reporters. That is not how it works now. Section 702 collects from American providers — meaning the collection point sits inside a commercial relationship the affected party entered voluntarily, disclosed in full, and agreed to in writing. Nothing has to be hidden anymore. Once the material is already pooled at a handful of reachable chokepoints, secrecy becomes optional; a certification suffices. That is the real innovation of the last twenty years, and it is why reform has become a permanent condition rather than an event. There is nothing left to expose.

Which puts the loop somewhere larger than surveillance. What sits at those chokepoints now is not a set of phone calls. It is closer to the whole recorded reach of human thought — what people write, ask, search, and confide — concentrated in a few private hands and available on subpoena and certification. The governance question for that pool is precisely the one this story ends on: who is in the room. And standing doctrine, which requires an identifiable individual with a particularized injury, is structurally incapable of recognizing a harm whose injured party is a collective. There is no plaintiff for an enclosure. That is the deepest version of the closed loop, and it is not a 2006 problem.

Prediction, offered without pleasure and without much uncertainty: the next ruling against a surveillance program will be legally correct, will arrive after the program has relocated, and will be vacated or mooted on procedural grounds having nothing to do with whether the conduct was lawful. Not because the judges are captured. Because the doctrine requires a plaintiff who knows, and the not-knowing is the product.

Judge Taylor died in 2017. Her line about hereditary kings is still quoted, mostly by people who do not mention that it was vacated eleven months later on a technicality and legalized eleven months after that. Twenty years on, the anniversary is the only part of this that recurs on schedule.