IETF DULT Specifies Behavior, Not Architecture

Vendor and Product Reality

The DULT Working Group emerged from a confluence of pressures. Apple's AirTag, launched in 2021, surfaced a category of harm — covert location tracking enabled by ubiquitous Bluetooth-based crowdsourced finding networks — that no single vendor could address alone, because the tracker emitted by one vendor's device travels through the detection surface of every other vendor's device. Apple and Google jointly published a draft specification in 2023 proposing a common advertising format and detector behavior; the IETF DULT Working Group formalized the standardization track, and the participant list grew to include the major tracker manufacturers and several civil-society contributors.

The specification's scope is precise. It defines how a tracker should advertise its presence in a way that any DULT-compliant detector can recognize. It defines how a DULT-compliant detector should behave when it identifies a tracker that appears to be moving with a person who is not the tracker's owner: alert thresholds, alert content, and escalation to platform-level UI. It defines how the affected individual can obtain identifying information sufficient to disable the tracker or, in escalated cases, to support a law-enforcement response. The result is a behavioral interoperability standard that allows any compliant tracker and any compliant detector to participate in a shared anti-stalking flow regardless of vendor.

The work is operationally important. It produces a baseline that did not exist three years ago, it commits major vendors to a shared protocol, and it gives civil-society organizations a stable surface against which to evaluate compliance. None of what follows diminishes that.

The Architectural Gap

DULT specifies behavior. It does not specify the architecture beneath the behavior, and the gap matters because the threat model DULT addresses is fundamentally architectural. Two structural properties of the current finding networks remain unchanged after DULT compliance, and both are load-bearing for the harms DULT is meant to mitigate.

The first is centralized network authority. The Apple Find My network, Google's Find My Device network, Tile's network, and Samsung's SmartThings Find each remain governed by a single operator who controls the credentialing of readers, the routing of location reports, and the policy under which a tracker becomes findable. A DULT-compliant tracker advertises across vendor boundaries, and a DULT-compliant detector recognizes it, but the underlying location reports continue to flow into the operator's infrastructure under the operator's policy. There is no architectural mechanism by which an authority outside the operator — a domestic-violence shelter operating a sweep, a court issuing a protective order, an airport security team responding to an incident — can participate in reader activation under credentials the network is structurally required to honor.

The second is opt-out detection. DULT places the burden of detection on the potential victim's device. The tracked individual must be carrying a DULT-compliant detector, that detector must be configured to alert, and the alert must be acted upon. For populations who do not own a recent smartphone, who share devices, or who are in coercive relationships where the device itself is not under their control, opt-out detection is structurally insufficient. The architecture provides no mechanism for protective detection to be activated on behalf of a person by a credentialed third party.

Together these gaps mean that DULT-compliant networks remain centralized authorities offering opt-out detection — improved over the pre-DULT baseline, but still architecturally identical to it. The behavior interoperates; the authority does not redistribute.

What the AQ Primitives Provide

The AQ semantic-discovery primitive describes a discovery substrate in which reader activation is governed by credentialing chains that can extend beyond the network operator. A reader can be activated under the operator's own credentials for ordinary lost-object recovery; it can also be activated under externally credentialed authorities — courts, regulators, accredited safety organizations — for activations the network is structurally required to honor under cross-recognition policies negotiated at the substrate level. The substrate itself, not any single operator, enforces the credentialing.

The n-party-coordination primitive describes how multiple authorities cooperate without merging. Apple, Google, Tile, Samsung, Chipolo, and Pebblebee remain independent operators with independent commercial and policy interests; the coordination primitive specifies how their networks recognize each other's credentialed activations, how disputes are resolved, and how an activation by one authority propagates to the readers operated by others. The result is structural anti-stalking: the architecture, not the vendor's discretion, determines that a court-issued sweep must be honored across all participating networks.

These primitives do not replace DULT. They sit above it. DULT continues to specify how a tracker advertises and how a detector behaves at the protocol level; the AQ primitives specify how reader activation is credentialed and how multi-party coordination is governed at the architectural level. The two layers compose cleanly because each addresses a distinct concern.

Composition Pathway With DULT

The composition is layered. At the bottom remain the DULT advertisement and detection behaviors, unchanged. A DULT-compliant tracker continues to advertise in the standardized format; a DULT-compliant detector continues to recognize and alert. Above that layer, the semantic-discovery primitive attaches a credentialing envelope to reader activations. A reader operated by an Apple device, a Google device, a Tile dongle, or a Samsung device participates in activations whose credentials the substrate validates and whose lineage the substrate records.

Above that, the n-party-coordination primitive specifies the cross-recognition policies that the participating networks negotiate. A protective-order activation issued by a credentialed judicial authority propagates to readers across all participating networks under terms the networks agreed to in advance. A safety-organization sweep operates under more constrained credentials but with the same architectural guarantee that participating networks honor the activation. The vendors retain control over what credentials they recognize and on what terms; the architecture ensures that once recognized, the activation is enforced uniformly.

For the specific harms DULT addresses, the composition closes the gap between behavioral compliance and structural protection. A tracked individual who does not own a DULT-compliant detector is no longer architecturally outside the protection regime; a credentialed third party can activate detection on their behalf. A network operator who would prefer not to participate in a particular activation class can negotiate that out of its cross-recognition policy, but cannot unilaterally void an activation it has agreed to recognize. The behavioral standard and the architectural primitive together produce a regime that neither could produce alone.

Commercial and Licensing Posture

Adaptive Query's posture toward the DULT participants — Apple, Google, Tile, Samsung, Chipolo, Pebblebee, and the broader Working Group — is explicitly non-competitive with the standards work. The patent describes the architectural layer above DULT, and licensing is structured to allow standards-track adoption: the primitives can be implemented by participating networks under licensing terms compatible with the IETF's standards-implementation expectations, and the cross-recognition policies the primitives require can be negotiated within the same multi-party governance forums that produced DULT.

For the participating vendors, the commercial value of adoption is in the regulatory and reputational dimension. Anti-stalking regulation is moving from voluntary best practices toward statutory requirements in multiple jurisdictions, and the requirements increasingly reach beyond protocol behavior into structural guarantees about who can activate detection and under what credentialing chain. A network whose architecture already supports credentialed third-party activation under enforced cross-recognition is positioned for those requirements; a network that has only DULT compliance is not.

The patent positions the primitive at the architectural layer where DULT and similar specifications are converging behaviorally but cannot, by their own scope, reach. The licensing pathway is designed to make that layer adoptable by the same vendors who built DULT, on terms compatible with the standards process they used to build it. The civil-society organizations that participated in the DULT process — domestic-violence advocacy groups, privacy non-profits, academic researchers studying intimate-partner surveillance — gain, in this composition, an architectural surface against which to credential their own protective activations rather than relying on per-vendor accommodation. That surface is what turns DULT's behavioral floor into a structural ceiling on the harms the working group set out to address.