A story flashed through the September news digest that everyone reads as routine: the UN chief for disarmament noted that not a single nuclear power has ratified the Comprehensive Nuclear-Test-Ban Treaty, and the US and Russia are openly threatening to return to testing. The moratorium, supposedly, holds on a gentleman's agreement. But hiding in this story is a non-obvious twist: the treaty is legally dead—it never entered into force in thirty years—yet the infrastructure built for it is alive and operating at full capacity. Hundreds of stations scattered across the planet, from Norway to Antarctica, listen to the Earth, oceans, and atmosphere around the clock—and it's precisely this "orphan without a law" that caught all six of North Korea's nuclear tests. The question I wanted to untangle: how did it happen that the most effective military verification system in history belongs to a treaty that formally doesn't exist?
The Comprehensive Nuclear-Test-Ban Treaty (CTBT) was adopted by the UN General Assembly in 1996—and has been stuck in a legal trap of its own design ever since. For it to enter into force, it needs ratification by all 44 states listed in Annex 2—those who possessed nuclear technologies at the time of signing. Thirty years later, nine of these forty-four still stand aside: China, Egypt, Iran, Israel, and the US signed but did not ratify; India, Pakistan, and the DPRK did not sign at all; Russia ratified in 2000, then withdrew its ratification in November 2023—Moscow explained this as "parity" with Washington, which never ratified the document.
As of 2026, of the 196 UN states, 188 have signed the treaty, 179 have ratified it—remarkable arithmetic for an agreement that is not in effect. The latest to join the signatory club in July 2026 was Tonga, whose nuclear ambitious past is limited, it seems, to cooking ovens.
Now here's the main thing. While lawyers have spent three decades arguing about the treaty's status, they managed to build what it was conceived for. The International Monitoring System (IMS) is 321 stations and 16 laboratories in 89 countries, from seismic sensors in mines to air sampling stations on tiny islands. About 90% of these 337 facilities are already operating in real time. The treaty hasn't entered into force—but the network that was supposed to be its eyes and ears has become one of the most reliable scientific-technical systems on the planet.
The IMS trick is multimodality. One and the same burst of energy in the Earth's core, ocean, or atmosphere leaves traces in four different media, and each duplicates the others.
Seismic stations—the system's backbone: 50 primary and 120 auxiliary, listening to ground vibrations. An underground explosion is easily distinguished from an earthquake by its wave "signature": the explosion has a dominant isotropic component, compressing rock in all directions at once. Hydroacoustic (11 stations) catch sound in the ocean—water conducts acoustics for thousands of kilometers, so an underwater explosion is heard from almost the other end of the planet. Infrasound (60 stations) monitor ultra-low frequencies in the atmosphere, inaudible to humans. Finally, radionuclide stations (80 plus 16 laboratories) search the air for radioactive particles and noble gases—xenon-133, which seeps through rock even from a carefully sealed mine and serves as legal "conclusive proof" of a nuclear explosion.
Data integrity is literally engineered into protection: the stream from each station has an embedded digital signature, and tamper sensors are mounted on equipment housings. Stations are operated by local institutes under contracts with the organization—a rare example of geopolitical infrastructure working on contractor trust worldwide.
How sensitive is the network? In the 1990s it was designed to detect 1 kiloton; according to CTBTO Executive Secretary Robert Floyd, reality exceeded the design—the system confidently sees explosions around 500 tons of TNT equivalent and below. The scientific estimate for the global seismic network threshold is about magnitude 4.0, which in favorable geology means hundreds of tons. For comparison: the Hiroshima bomb was 15 kilotons.
The best way to evaluate the system is to look at how it handled the only country actually testing nuclear weapons in the 21st century. All six North Korean tests from 2006–2017 were caught, and the evolution of this work is a mini-history of the entire network's maturation.
In 2006, at the first test, only 10 of 40 planned noble gas stations were measuring—in test mode. Nevertheless, the Yellowknife station (Canada, thousands of kilometers away) caught xenon-133 two weeks later, and wave analysis allowed them to designate a possible inspection zone of less than 1000 km²—the treaty's formal threshold.
By 2013, the system was already running at adult speed: the February 12 event at magnitude 4.9 was registered by 94 seismic stations and two infrasound stations within seconds, the revised bulletin went to states ahead of the contractual 48-hour deadline, the error ellipse narrowed to ±8 km. And 70 days later, the Takasaki station in Japan—about a thousand kilometers from the test site—caught xenon-131m and xenon-133, confirming the source was nuclear. In the arms control specialist community after this, people started talking about the CTBT being "effectively verifiable."
The apotheosis—September 2017. The explosion on September 3 at 03:30 UTC with magnitude 6.1 was caught by hundreds of stations worldwide within seconds; 41 primary and 90 auxiliary seismic stations, two hydroacoustic and one infrasound participated in the analysis. Data reached Norway in 11 minutes—7360 kilometers through the planet's core. The automatic verdict "man-made explosion" was ready in an hour, localization accuracy ±6.7 km, several times better than the treaty requirement. The second signal 8.5 minutes later turned out not to be a second explosion but a cavity collapse in Mount Mantap—the system distinguished even this.
Yield estimates diverged—and this is a separate honest lesson about the method's limits: the Norwegian Seismic Array NORSAR gave 250 kilotons, American authors up to 400, others around 120. The reason for the spread is banal: the empirical formula "magnitude → yield" requires a calibration constant for the specific test site, and no one conducted reliable reference explosions at Punggye-ri. Seismology sees energy, but converting it to kilotons is always an estimate, not a measurement.
No reliable xenon trace was found after the 2017 test—and this is not a failure but a pure detective story: September xenon detections in the region didn't match transport trajectories from Punggye-ri, but October readings from station RN58 near Ussuriysk pointed to delayed leakage from the test site. The explosion partially "shook" the rock mountain—weak aftershocks continued for months.
The advertising says "we'll detect anything anywhere"—the truth is more complex. The first blind zone is ultra-low yields: hydronuclear tests in metal chambers with practically zero energy output (per Daryl Kimball's estimate from the Arms Control Association) the system "doesn't see" in principle. The second is decoupling: an explosion in a pre-excavated cavity in rock salt dampens the seismic signal by about 70 times (the classic Sterling 1968 experiment), meaning a kilotonne can be masked to "audibility" of hundreds of tons. But here's a saving "but" hiding: salt is not suitable everywhere, the cavity needs months to build—and all this is visible from satellites. You have to hide not just the explosion but all its preparation.
The third zone is subcritical experiments: no chain reaction starts, no explosion occurs, and the treaty explicitly permits them. It's precisely this loophole the US Energy Secretary now points to, promising "tests without nuclear explosions." Formally—honest; in fact—the line between verifying arsenal reliability and rehearsing a real test becomes increasingly blurred.
Finally, radionuclide verification depends on wind and half-life: xenon-133 lives five days, and if leakage is delayed by weeks, chances of catching it evaporate. The event remains "probably nuclear" by seismics, but without legally clean proof.
The most interesting thing in this story is not engineering but political economy. The IMS has been operating with "preparatory commission" status for three decades already, because entry into force requires ratifications that don't exist and aren't foreseen. It's financed by member state contributions—the 2025 budget exceeded $139 million—and no one seriously raises the question of disbanding it, though formally it serves an unratified treaty.
So why does everyone tolerate this? Because the system benefits everyone simultaneously—the rarest case of geopolitical symmetry. Open access to all data is guaranteed to every signatory state: a small country without a single satellite gets the same intelligence stream about nuclear explosions as a superpower. China and India would gain more from resuming tests than the US and Russia (who have thousands of historical tests and accumulated statistics), so their motivation to break the system is lowest. The network watches all known test sites—Punggye-ri, Lop Nur, Novaya Zemlya, Nevada—around the clock, and any suspicions of a "hidden test" are immediately checked against its data: this happened with American accusations against China regarding the Lop Nur explosion in 2020.
And incidentally, the system built for fear turned out useful for life: 164 seismic and hydroacoustic stations feed tsunami warning centers in 23 countries in real time (the latest agreement—Mexico, August 2026), gaining up to three minutes over alternative sources. After the 2004 tsunami that claimed over 230,000 lives, this cooperation became permanent.
The formula into which Floyd himself packed the situation in conversation with UN correspondents in September 2026 sounds almost engineering: his task is to give states "confidence that any explosion anywhere and anytime will be noticed." The treaty bans nuclear testing, but the real work is done not by it but by the network, which guarantees one thing: hiding an explosion from the world is impossible, which means any violation becomes common knowledge the same day. A moratorium that holds on calculation rather than law is still calculation with an audience. And the audience has been sitting with perfectly tuned headphones for thirty years now.