A government fell in Riga on the fourteenth of May 2026, exactly one week after the event that made it fall. On the seventh of May, drones came down near a fuel depot outside Rezekne in two separate strikes, damaging storage infrastructure and injuring no one directly. By Ukraine's own account afterward, the drones were Ukrainian, diverted off course by Russian electronic warfare during strikes elsewhere in the war next door. Latvia was not attacked by an enemy that week. It was struck by a friendly state's own weapon, bent off its intended path by an adversary's jamming, and the political fallout took seven days to run its course: Defense Minister Andris Spruds resigned on the tenth, coalition partner the Progressives withdrew their support on the thirteenth, and Prime Minister Evika Silina resigned the following day.

Friendly fire still ends governments.

That is a stranger place for a defensive alliance to arrive at than an unidentified attacker would have been. Nobody in Brussels held a summit and decided that a government's survival should now hinge on whether an ally's drone can be steered straight through someone else's jamming. It happened anyway, and it took a full week even to establish that much and let the argument run its course. The intercepts this piece is actually about get seconds, not a week. That gap alone is one data point in a pattern that has been building for sixteen months along NATO's eastern edge, a pattern this piece traces from a cut cable in the Baltic to a downed drone over Lithuania this month.

What Article 5 Actually Promises

NATO's fifth article is usually summarized as an automatic tripwire, an attack on one is an attack on all. The actual mechanism is narrower, and in one specific way more permissive, than that summary suggests. The treaty commits each member to take, individually and in concert with the other parties, "such action as it deems necessary, including the use of armed force," to restore and maintain the security of the North Atlantic area. NATO's own current legal guidance is direct about what that means in practice: the determination of whether an armed attack has occurred can be made individually or collectively, each ally decides for itself how to fulfil its obligation, and NATO's coordination role does not preclude unilateral or bilateral action by a member defending itself. Article 5 has been formally invoked exactly once, after September 2001, and what it produced was NATO AWACS aircraft patrolling American skies under Operation Eagle Assist, a demonstration of collective solidarity rather than a war-fighting decision in its own right. It was never built to require the alliance's permission before a member state defends its own airspace against an incoming object. That right exists independently of the treaty, and Article 5 says plainly that it builds on that right rather than replacing it.

Article 5 was never the mechanism standing between a single country and a decision to fire. It was always something narrower: a promise of collective political and military weight once a member decides, alone or together with the others, that an armed attack has actually occurred.

NATO built a separate mechanism for exactly this level of incident, and it has been used far more than Article 5 ever has. Article 4 lets any member call the North Atlantic Council into consultation whenever it judges its territorial integrity, political independence, or security threatened, without requiring anyone to agree an armed attack occurred at all. It has been invoked nine times since 1949, mostly by Turkey, and twice inside the same two-week span this piece is built around: Poland on the tenth of September 2025, the same day it disclosed the drone incursion that triggered NATO's first intercept, and Estonia on the twenty-third, after a Russian fighter jet violation four days earlier. Both consultations happened, on schedule, exactly as the treaty intends. Neither one decided whether to fire. That decision had already been made and executed, in seconds, well before the ambassadors in Brussels sat down to discuss it.

That distinction matters, because it means the discontinuity this piece is tracking is not the one it might first appear to be. Nobody stole a decision that used to require thirty-two votes. What happened instead is that an entire category of lethal decision, one that used to be rare enough to need no standing procedure at all, has become routine at a level Article 5 was never designed to reach in time to shape.

The Cable, the Airspace, and the Sentry

The sequence has an identifiable start date. On the fourteenth of January 2025, Secretary General Mark Rutte announced Baltic Sentry, after undersea cables connecting Lithuania to Sweden, Germany to Finland, and Estonia to Finland were damaged within the space of two months. Rutte's language was already the language of enforcement: ship captains, he said, "must understand that potential threats to our infrastructure will have consequences, including possible boarding, impounding, and arrest." He was careful, in the same appearance, to draw a line NATO has kept ever since: law enforcement itself remains a national responsibility. NATO's own description of Baltic Sentry frames it as an alliance-wide surveillance and deterrence layer, run under the Supreme Allied Commander, feeding warning and coordination to the national authorities who hold the actual legal power to stop, board, or arrest a vessel. Baltic Sentry added ships, maritime patrol aircraft, submarines, satellites, and a new fleet of naval drones to the Baltic. It did not add a NATO power to board a ship. It added a shared picture, built once, for thirty-two navies to act on separately.

Eight months later the same design repeated in the air. Overnight between the ninth and tenth of September 2025, roughly twenty Russian drones crossed into Polish airspace, and NATO aircraft shot down four of them, the first time the alliance had done so. Two days later, on the twelfth, Rutte and the Supreme Allied Commander, General Alexus Grynkewich, announced Eastern Sentry, a reinforced air and ground posture along the alliance's eastern flank, with assets from Denmark, France, the United Kingdom, Germany, and other members joining in the following weeks. Grynkewich's own framing was explicit about scope: "Although the immediacy of our focus is on Poland, this situation transcends the borders of one nation." The posture generalized before any single incident had even been resolved.

Both operations built the same thing: a shared NATO layer of detection and warning, sitting above a patchwork of national authorities who retain the actual power to act. Neither invoked Article 5. Neither needed to. What they created was infrastructure for a decision that, until 2025, had rarely needed infrastructure at all, and both were announced by the same man, in the same voice, eight months apart.

The Machine Was Already Running

The shared picture this piece keeps naming has an actual address, in fact three of them: Uedem in Germany, Torrejón in Spain, and, since October 2025, Bodø in Norway, NATO's network of Combined Air Operations Centres under Allied Air Command at Ramstein, built around the NATO Integrated Air and Missile Defence System, NATINAMDS for short. Sensor posts across European NATO airspace feed Recognized Air Pictures into this network of three centres around the clock, overlapping by design for redundancy, tracking up to thirty thousand air movements a day, and cueing interceptor aircraft toward anything that picture cannot yet classify. This is the machine that told a Romanian pilot which drone over Estonia to engage, visually identified before the shot, well before either government involved could establish who had launched it, why it had crossed the border, or what political responsibility followed.

NATINAMDS is not a purely post-Cold-War invention. Its lineage runs to NATINADS, approved by NATO's Military Committee in December 1955, and NADGE, its ground environment, completed by 1962, automating a command relationship that was already two years old: from 1960, member states had placed their air defence forces under the Supreme Allied Commander Europe in the event of war. What changed across seven decades was the object the system had to classify, Soviet bombers, then Cold War fighters and missiles, now drones bent off course by electronic warfare. The command relationship itself never changed.

None of this was built for 2026. Baltic Air Policing, the mission running that Estonian intercept, began on the twenty-ninth of March 2004, a Belgian F-16 landing at Šiauliai Air Base in Lithuania because Estonia, Latvia, and Lithuania had joined NATO with no combat jets of their own. A second base opened at Ämari in 2014, the year of Crimea. More layers were added in 2022, the year of the full invasion. For two decades the mission's job stayed identification and escort, not destruction.

Baltic Air Policing is not the exception either. NATO Air Policing as a whole dates to 1961, run continuously through the Cold War and after, and it now protects several allies that do not maintain a full national fighter-policing capability of their own, from Iceland to Montenegro, each patrolled by a neighbor's air force under the same standing arrangement. Multiple NATO members fly someone else's jets over their own territory as a matter of permanent institutional design, not emergency improvisation. The eastern border did not invent the arrangement now absorbing its drones. It only made an arrangement built for peacetime deterrence newsworthy again.

This piece has been describing an architecture that accumulated. The mission running the sharpest edge of it did not accumulate. It was already there, twenty-two years old, waiting for a rule change large enough to let it fire.

That rule change has a date. On the eighth of July 2026, Lithuanian President Gitanas Nausėda announced that Baltic Air Policing itself had been upgraded from air policing to air defence, giving its pilots standing authority to destroy threatening objects rather than only identify and escort them. His own framing conceded the point directly: "what is happening today is not a totally peaceful environment." The upgrade arrived only after Baltic Sentry in January 2025, the first alliance-wide shootdown over Poland that September, and a lengthening run of national intercepts through the spring and summer of 2026 had already made the old identify-only rule look obsolete in practice, well before NATO changed it on paper.

That sequence is the tell for a fair objection this piece needs to take seriously: if NATO deliberately upgraded an existing mission, in daylight, with a named decision and a named date, is this really an unplanned migration at all, or a design choice this reading is mistaking for drift? The timeline answers it. NATO did not sit down in early 2025 and design a faster, more distributed air defense architecture for its eastern flank. It spent eighteen months catching its own rulebook up to an operational reality that had already moved past it, then formalized the change in a single announcement once the gap became too obvious to leave alone. The paperwork followed the pattern. It did not create it.

Shared Detection, Distributed Trigger Authority

Look closely at who actually pulled the trigger in each of the specific incidents that followed, and the picture is not five capitals quietly reclaiming their own war powers. It is something more tangled, and more revealing.

Romania passed the law behind that authority more than a year earlier, in the spring of 2025, when interim president Ilie Bolojan signed a decree on the nineteenth of May authorizing the military to neutralize or destroy unauthorized drones over Romanian territory in peacetime, a category of authorization that had simply not existed before. The law then sat mostly unused for over a year. Only on the twenty-fourth of July 2026 did a Romanian F-16 finally invoke it over Romanian territory, the first of four such intercepts credited to Romanian forces within weeks. The gap between the law and its first use is itself a data point: the authority existed well before the operational tempo caught up to it, the same pattern this piece keeps finding in Vilnius and at Baltic Air Policing headquarters. That is still a clean national case once it does arrive: national law, national territory, national aircraft. Romania's own defense ministry later put a number on the acceleration behind that law: at least twenty-three drone violations of Romanian airspace recorded in 2026 alone by the twenty-first of August, against eighteen recorded across the four full years from 2022 through 2025 combined.

The next case is not. On the nineteenth of May, a drone over Estonian territory was destroyed not by an Estonian aircraft acting alone but by a Romanian F-16 flying under NATO's Baltic Air Policing mission, with the engagement directed by Latvia's Control and Reporting Center, itself one node feeding the same NATINAMDS picture running out of Uedem. One country's airspace, another country's jet, a third country's command center, and a target that Estonia's own military says was visually identified before the shot was fired. Latvia's own intercepts, on the eighth of June and the fourteenth of August, followed the same pattern, allied NATO Air Policing fighters operating alongside national air defense rather than a single national chain of command acting in isolation. Lithuania's newest law, passed in September 2026, authorizes national action, but the drone its parliament's new statute was written to cover was destroyed on the fifteenth of that same month by Italian fighters flying under NATO's own Baltic Air Policing mission, the wreckage found near the village of Pratkūnai, roughly sixty miles west of Vilnius. Lithuania's own air force commander and its national crisis center both confirmed the intercept ran through that same allied, not national, chain. Rutte called the strike proof of Putin's "desperation and also his desire to sow fear and terror," a day after warning that strikes near NATO territory would only deepen the alliance's support for Ukraine.

What migrated is not the authority to decide. What migrated is the level at which any decision gets made at all: a shared NATO detection and command layer, feeding national legal authorizations, executed by whichever allied aircraft happens to be airborne, in a decision window measured in seconds rather than the weeks a formal attribution would take. That is a harder thing to reverse than a national power grab would be, because there is no single actor left to hand the decision back to.

The Wall That Classifies Faster Than It Attributes

Running beneath all of this is an infrastructure project meant to make detection, if not attribution, routine. The European Drone Defence Initiative, the so-called drone wall, began deployment in the first quarter of 2026 along the European Union's roughly four-thousand-kilometer eastern border, with initial capacity targeted for the end of that year and full functionality for the end of 2027, according to the EU's own roadmap. Its own description is deliberately not a single fence but a layered, interoperable network, radar and acoustic and radio-frequency sensors feeding the picture, electronic warfare tools meant to jam or spoof an incoming drone's guidance, and interceptor systems, some kinetic and some not, to finish what the sensors start. One cost estimate, published by the Austrian defense outlet Militär Aktuell and attributed to a planning paper from a country on the EU's eastern border, puts the sensor layer alone at close to nineteen million euros per hundred kilometers of border, with twenty to forty percent of deployed devices expected to need replacement every thirty days under operational conditions, a national planning estimate rather than a published EU design specification.

The initiative leans on pooled EU funding, principally the Security Action for Europe instrument and the European Defence Industry Programme, and its own design brief is explicit that procurement will align with existing NATO command structures rather than build a parallel one. The wall is not competing with NATINAMDS. It is wiring itself into it.

The wall's premise is classification, not attribution. Modern radar and AI-supported sensor networks are built to sort an incoming object by type and trajectory quickly, and the Estonian case shows the system working exactly as designed: the drone was identified before the shot was fired. What the wall was never built to establish, in the seconds it has to work with, is who sent the object and why, the slower and harder judgment that used to be the entire point of the decision architecture built around Article 5.

The wall can classify an object faster than Europe can attribute an act. That gap, not any missing radar capability, is the actual discontinuity this border now runs on.

The Loop

A command system built in 1955 to track Soviet bombers is repurposed to track everything else. Cables are cut in the Baltic. NATO builds a shared detection layer and calls it Baltic Sentry. Drones cross into Poland. NATO generalizes that layer along the whole eastern flank and calls it Eastern Sentry. A twenty-two-year-old air-policing mission gets a new rule of engagement in July. Incidents continue weekly through 2026. National legal authority to fire expands, country by country, while the aircraft that actually pull the trigger increasingly belong to whichever ally is airborne under NATO's own air-policing command. A shared picture feeds a national law. A national law feeds a shared picture. A continent-length sensor wall is built to classify the next object in seconds, long before anyone can attribute it. Taken together, those decisions now function as a machine: seventy years of command architecture operating at a tempo its political layer was never built to match.

The Strongest Objection

The strongest objection to this reading does not dispute a single date or intercept. It says that nothing described here is actually new. Air forces have always shot down hostile or unidentified aircraft crossing their own borders on their own military's judgment, in seconds, without reference to any multinational council, because that is what an air-defense system is for. NATO air forces made exactly this kind of unilateral, seconds-long judgment throughout the Cold War, under Quick Reaction Alert postures that predate every incident this piece describes by decades. NATO's collective machinery was never built to authorize individual intercepts. It was built for a much larger and rarer question, whether a sustained and attributable campaign amounts to an armed attack requiring a collective response, and nothing in the pattern above shows that threshold has been crossed, eroded, or even approached. Reading a permanent, unremarkable feature of national air defense as a structural transformation of the alliance, on this view, mistakes ordinary vigilance for a discontinuity that was never there.

A sharper version of the same objection would point straight at Article 4. NATO invoked it twice within two weeks in September 2025, exactly when this piece says the pattern began, which shows the alliance already had a working channel for incidents below the Article 5 threshold and used it precisely as designed. On that view there is no gap here for a piece like this one to discover. The consultative machinery worked as intended, and the intercepts that happened alongside it were never meant to wait for that machinery to begin with.

This counterargument is serious, and it is correct about what has not happened: no Article 5 event has occurred, no collective decision has been overridden, and the North Atlantic Council's authority over a genuine armed attack remains exactly what it was in 1949. It is also right that Article 4 worked as designed. What it does not explain is why a functioning consultative channel and a functioning operational one, running side by side, produced no forum in which anyone treats the accumulation itself as the object of decision. The reading offered here does not claim that authority has eroded. It claims something narrower and, in a sense, more mundane: that the volume and tempo of incidents sitting below that threshold have grown large enough, one member state alone recording more airspace violations in under eight months of 2026 than in the four preceding years combined, across several countries rather than one, that an alliance built to handle the rare case now runs a permanent, distributed operation handling the common one, with no forum in which anyone examines the pattern as a pattern rather than one intercept at a time. Ordinary vigilance, repeated enough times, in enough places, becomes an architecture. Nobody designed that architecture as a whole. It simply accumulated, incident by incident, in exactly the space this piece describes.

What would prove this reading wrong is specific enough to name. If the incident rate along NATO's eastern border falls back toward its 2022 through 2025 baseline and stays there, and if the July 2026 upgrade to Baltic Air Policing and the national laws passed alongside it turn out to be a one-time correction rather than the start of a durable operating layer, the claim that decision-making has migrated to a permanent sub-threshold tier will not survive contact with what happens next. A single incident later reaching an actual Article 5 determination would not, on its own, disprove this reading. It would only show that the threshold above the layer still functions, which this piece has never disputed. Nothing in the record so far points that way, but naming the condition that would change it is part of the discipline this piece is trying to hold itself to, and it is a condition anyone can check without taking this piece's word for any of it.

What the Architecture Does Not Account For

Go back to Riga, and to the cabinet that fell over a drone that was not even hostile. That is the detail worth sitting with. The new problem this border has produced is not that Article 5 has quietly disappeared, or that some collective decision was stolen out from under the North Atlantic Council. The problem is narrower, and in its own way more unsettling: a growing share of the lethal encounters happening along NATO's eastern edge now sit below the threshold Article 5 was ever built to reach, decided in the seconds between a radar contact and a pilot's order to fire, by whichever allied aircraft happens to be airborne, under a shared detection architecture that can classify an object far faster than any government can attribute it.

Nobody voted for that migration, and nobody needed to. It was never a decision in the sense Article 5 governs. It was an accumulation of separate, locally reasonable choices, a cable-protection mission, an air-policing mission, a Romanian law, a Lithuanian law, a shared radar picture, that has produced a permanent operating condition nobody designed and nobody is now positioned to review as a whole.

The architecture accounts for the cable, the drone, the sensor, and the statute. It even has a building for the decision, Uedem, Torrejón, or Bodø, wherever the Recognized Air Picture is being watched at the moment an unknown track appears. What it has no line for is the person standing beneath whatever the command chain watching that picture decides, in the seconds available, to treat as a threat. NATO still decides whether to go to war together, exactly as it did in 1949. The border it is defending no longer waits for that decision to be relevant.

Sources & Documents

This piece rests on NATO's own announcement of Baltic Sentry, NATO's own announcement of Eastern Sentry, and the full transcript of Rutte and Grynkewich's joint press conference announcing it, alongside Allied Air Command's own account of twenty years of Baltic Air Policing, Allied Air Command's own page on its Combined Air Operations Centres, which is the source for the thirty-thousand-movements figure and its European-airspace scope, NATO's own current page on NATO Integrated Air and Missile Defence, and The Aviationist's report on the October 2025 opening of the third Combined Air Operations Centre at Bodø. The Latvian government crisis is sourced to Latvian public broadcaster LSM's reporting on the resignation and Al Jazeera's coverage of the same event. The Romanian shoot-down law is sourced to Romania Insider's report on the May 2025 presidential decree and its report on the July 2026 first use of that authority. The September 2026 Lithuanian intercept is sourced to NPR's own reporting. The European Defence Industry Programme's funding structure is sourced to the Council of the European Union's own press release on its final approval. The sensor-network cost estimate in the wall section is sourced to Militär Aktuell's reporting on a national planning paper, not a published EU design specification, and is presented as such.

Jerry van der Laan writes The Manifest Archive, a forensic study of the institutions, incentives, and infrastructures that determine outcomes long before the public debate about them begins.