The Race Beyond GPS Needs A Rule For When Machines Stop Trusting Time – OpEd

Key Takeaways:

• Late-August and September trials: Q-CTRL GPS-free Coral Sea run with quantum gravimetry; Royal Navy, Saab and Aquark cold-atom clocks on military radars with planted timing errors; DIU–Honeywell magnetic navigation over the Pacific for 4+ hours without GPS. Finnish authorities said four Ukrainian military drones found in Finland were thrown off by strong GNSS jamming plus wind. EASA is logging GNSS outages from the Baltic and Eastern Europe to Türkiye, the Middle East and Asia.
• Clock, inertial, gravity, and magnetic sensors are not the same thing as a position. A good holdover clock is timing, not autonomous nav. GPS going dark is the easy case; a live but wrong, stale, or inconsistent reference is harder. NIST’s 2026 PNT profile draft asks how AI and third-party data change that risk. A system should know its source, age, holdover, growing error, and whether independent references agree. Decimal places are not trust.
• Rule: as confidence falls, permissions fall—wider margins, a second source, less automation, or hold high-consequence acts. AI confidence does not fix a bad clock. Layer SATNAV, local clocks, INS, maps, and MagNav/gravity; count independent chains, not logos. Allies need a shared vocabulary for time quality and degraded modes, not identical hardware. Require a PNT trust envelope and drill wrong-but-plausible signals plus restoration—not only total loss.

Quantum clocks, magnetic maps and gravity sensors are moving out of the laboratory as GNSS jamming spreads across transport and security systems. But resilient navigation requires more than another source: machines must know when timing and position evidence have become too uncertain for the action they are about to take.

The race to operate without GPS has accelerated from several directions at once. In late August, Australian company Q-CTRL reported a GPS-free maritime trial in the Coral Sea using quantum gravimetry to correct inertial navigation. In Britain, the Royal Navy, Saab and Aquark tested cold-atom clocks with military radars and deliberately introduced timing errors to see whether the network could recognize them. And earlier this month, the U.S. Defense Innovation Unit and Honeywell flight-tested magnetic navigation over the Pacific for more than four hours without relying on GPS.

Those demonstrations are arriving because the problem is no longer theoretical. Finnish authorities said last week that four Ukrainian military drones found in Finland earlier this year had been diverted from their intended routes by strong GNSS interference combined with wind. The European Union Aviation Safety Agency’s current GNSS disruption reporting has also tracked outages or alterations across flight-information regions stretching from the Baltic and Eastern Europe to Türkiye, the Middle East and Asia.

That makes alternative navigation strategically important. It also creates a risk of its own: governments may start treating every impressive new clock, inertial sensor or quantum device as a replacement for the thing GPS used to provide.

It is not.

A resilient positioning, navigation and timing system needs more than a better backup source. It needs an explicit rule for what the machine may still do when its confidence in time or position begins to decay.

A clock is not a position

The first mistake is to collapse positioning, navigation and timing into a single capability. They are related, but they are not interchangeable.

An atomic clock can provide a highly stable local time reference. An inertial sensor measures motion. A magnetic or gravity-matching system can compare measurements with mapped features of the Earth. A usable navigation solution emerges only when those functions are combined with estimation, calibration, initialization and other sensors.

That distinction matters for procurement. A clock that performs extremely well in holdover has demonstrated a timing capability. It has not, by itself, demonstrated autonomous positioning. A quantum gravimeter that constrains inertial drift at sea has demonstrated something different from a radar network whose timestamps remain coherent during deliberate signal manipulation.

The current wave of trials is valuable precisely because it is beginning to separate those claims. The policy error would be to put them back together under a single political label such as ‘quantum navigation.’

The missing variable is trust state

GPS failure is the easy case to understand because the signal disappears. The more difficult case is a reference that remains available but is wrong, stale or inconsistent with another source.

A machine can receive a valid instruction at the wrong time. It can also combine two correct observations as though they were simultaneous when their clocks disagree. A position estimate can look precise while the assumptions that keep its error bounded are no longer true.

That is why timing quality should be treated as operational evidence rather than hidden infrastructure.

NIST has documented how critical sectors use GPS-derived time for telecommunications, finance and electric power. Its 2026 draft update to the Foundational PNT Profile also asks how emerging technologies including AI affect PNT use and whether third-party and data-dependency risks are adequately addressed.

The next step is to make those dependencies machine-readable.

A consequential system should know which source it is using, how old that evidence is, whether it is operating in holdover, how much uncertainty has accumulated, whether independent references agree, and which functions actually depend on the level of precision being claimed.

A timestamp should not be treated as trustworthy merely because it contains many decimal places.

Authority should contract with uncertainty

This is where PNT resilience becomes a governance problem.

Different actions need different levels of timing and navigation confidence. A record written for later audit may tolerate more timing uncertainty than a synchronized control action. A vessel continuing on a broad safe route may tolerate more position uncertainty than a close-proximity maneuver. A telecom network, power-grid application, trading system or autonomous vehicle will have its own thresholds.

The rule should therefore not be ‘GPS lost, stop everything.’ Nor should it be ‘backup available, continue normally.’

The better rule is: when the evidence supporting a function weakens, the permissions attached to that function should change with it.

Some systems may widen operating margins. Others may require a second source, reduce automation, enter a restricted mode or wait for fresh confirmation. Local protective functions can remain available even when higher-consequence actions are temporarily withheld.

This is especially important for AI-enabled systems. An AI system can remain highly confident about a recommendation while the timing relationship among its inputs has become unreliable. Confidence in the model cannot repair uncertainty in the clock.

Quantum technology does not remove the integration problem

The recent demonstrations should encourage investment, not hype.

The U.S. MagNav flight showed how magnetic-field measurements can improve positioning without GPS. Q-CTRL’s maritime trial showed that gravity sensing can help bound inertial drift. The Royal Navy trial showed how cold-atom timing can support a network when conventional navigation timing is manipulated.

Each result answers a different question. None means every ship, aircraft or infrastructure operator should replace GPS with one new box.

The real architecture will be layered: satellite navigation when available, local clocks, inertial sensors, magnetic or gravity references, terrestrial distribution, maps and conventional instruments. The hard part is not simply collecting more sources. It is knowing when those sources are genuinely independent and what to do when they disagree.

Two clocks are not redundant if both inherit the same upstream reference. Two navigation systems are not independent if they share the same calibration failure. Resilience should be measured through dependency chains and tested failure modes, not by counting logos on a procurement slide.

The international problem is confidence, not hardware

No country will build this stack in exactly the same way. That is not a problem.

The United States may emphasize magnetic navigation and resilient timing. Britain is testing sovereign cold-atom clocks. Australia is pushing quantum gravimetry. European civil aviation is confronting widespread GNSS disruption while operators across the Middle East and Asia face their own interference environments.

What partners need in common is not identical hardware. They need a shared operational vocabulary for time quality, position uncertainty, source provenance and degraded-state behavior.

An allied aircraft, commercial vessel, autonomous system or cross-border infrastructure service should be able to say more than ‘navigation available.’ It should be able to say how trustworthy the estimate is, what supports it and which actions remain inside the justified envelope.

That can be done without revealing sensitive sensor designs. Confidence can be interoperable even when technology is not.

Build the trust envelope before the next outage

Governments and infrastructure operators should require a PNT trust envelope for consequential systems.

The envelope would identify the timing and position evidence required for different classes of action, define how uncertainty grows during holdover or sensor loss, specify the behavior when sources disagree, and state what evidence is required before full operation returns.

Exercises should include plausible but wrong references, not only total signal loss. Restoration should be tested too: a returning source should not regain operational influence simply because the receiver can see it again.

This is a narrower proposal than replacing GPS, and in practice it is more important.

The world may eventually deploy remarkable quantum clocks, gravimeters and magnetic sensors at scale. They will still age, drift, disagree, depend on maps and interfaces, and operate inside systems whose consequences vary by mission.

The goal is not a clock that never lies.

It is a machine that knows when time has stopped being good enough to justify what it wants to do.