Vendor and Product Reality

Stellantis was formed in 2021 from the merger of Fiat Chrysler Automobiles and Groupe PSA. Its stated technology consolidation strategy rests on three software-defined platforms: STLA Brain (the central computing and service-oriented architecture), STLA SmartCockpit (the cabin and infotainment layer, developed with Foxconn under the Mobile Drive joint venture), and STLA AutoDrive (the autonomous-driving stack). STLA Brain is the substrate on which AutoDrive runs: a high-bandwidth in-vehicle network, over-the-air update capability, and a domain-controller architecture that consolidates ECUs into a smaller number of high-compute nodes.

STLA AutoDrive has been publicly described by Stellantis as supporting Level 2, Level 2+, and ultimately Level 3 hands-free, eyes-off driving. The Jeep Wagoneer S has been publicly associated with early AutoDrive deployment, with premium Stellantis brands including Maserati expected to follow. The competitive frame is well-defined: GM Super Cruise, Ford BlueCruise, Mercedes Drive Pilot (among the first SAE Level 3 systems certified for public roads in the United States, in Nevada and California), Tesla's Full Self-Driving suite, and the various Chinese-market systems from BYD, NIO, and Xpeng. Mercedes Drive Pilot is a public demonstration that Level 3 can be certified in the United States; the certified operational design domain is also, per Mercedes' own published conditions, narrow, bounded by speed, road type, weather, and lead-vehicle presence.

Stellantis' challenge is not building a Level-3 stack. It is building one that can be certified, deployed, and updated across fourteen brands and four regulatory regions (UNECE in Europe, NHTSA/FMVSS in the United States, GB standards in China, and a fragmented set in the rest of the world) without each brand-region-feature combination becoming a bespoke compliance project.

The Architectural Gap

Hands-free driving is not, architecturally, a single feature. It is a continuous negotiation between graduated autonomy modes, driver-engaged, hands-off-eyes-on, hands-off-eyes-off, minimum-risk-maneuver, under conditions that change second by second. The transitions between these modes are exactly the events regulators care about: when did the vehicle decide it could accept eyes-off operation, on what evidence, under whose authority, and what is the auditable record of the transition. Conventional automotive software architectures represent autonomy mode as a state variable in the driving stack. As an architectural category, they generally lack a first-class structural representation of commitment as a composite-admissibility, multi-authority transition with a cryptographically continuous, lineage-recorded record. This is the axis this analysis addresses, not a claim about the internal implementation details of STLA Brain, which Stellantis has not fully published.

The multi-authority dimension is sharper at Stellantis than at any single-brand peer. A given Level-3 transition must be admissible under the homologation envelope of the destination market, the operational design domain declared by Stellantis to that regulator, the brand-specific feature set sold to the customer, the regional insurance and liability framework, and the driver's contractual acceptance of the system. These authorities do not align across brands or regions. A Maserati in Germany, a Jeep in Michigan, a Citroën in France, and a Dodge in Ontario each operate inside a different intersection of governing authorities. Without an architectural substrate that composes these authorities at runtime, every brand-region launch becomes a separate compliance and software project, and every regulatory update becomes a fleet-wide regression.

What the AQ Governed Actuation Primitive Provides

The governed actuation primitive treats commitment as graduated and authority-composed. Each autonomy mode is a stage with an explicit admissibility predicate; transitions are events that require the composed assent of the authorities entitled to govern them; ratification is a structural step, not a log entry. Applied to a vehicle, the primitive expresses Level 2, Level 2+, Level 3 within an operational design domain, and minimum-risk-maneuver as distinct stages, each with its own admissibility envelope and its own authority composition. A vehicle does not "engage Level 3"; it executes a stage transition whose admissibility was checked against the live composition of regulator, manufacturer, brand, region, and driver authorities, and whose record is part of the substrate of the action itself.

Concretely, a proposed mode transition is passed to a composite admissibility evaluator that reads authority-credentialed observations, an authority taxonomy, observation freshness, and governance policy, and returns not a binary permit/deny but one of six outcomes: admit, gate (permit subject to added constraints), defer (hold pending corroboration, with a deferral-expiration parameter), solicit (emit a governed discovery query for additional evidence), reject (with a rejection-reason classification), or escalate (raise a cross-domain condition to a higher authority). A skilled implementer can build this: the evaluator is a policy-parameterized function over a credentialed observation set and a configurable authority taxonomy that supports arbitrary depth plus dynamic escalation and de-escalation, each with a maximum duration and geographic or logical scope. A reversibility-aware commitment-point evaluator classifies each proposed actuation into reversibility classes (fully reversible, partially reversible, irreversible), elevates admissibility thresholds as the actuation approaches its point of no return, and prefers reversible paths where two are admissible. An emergency-preemption mechanism enforces per-authority preemption budgets with expiration so that a higher authority can preempt a mode without unbounded override. Every evaluation, mode selection, preemption event, and commitment-point transit is written to a lineage field, and the system degrades gracefully to more conservative modes under reduced confidence rather than failing to a binary. These mechanisms, and variations across signaling medium, authority taxonomy, aggregation topology (distributed, centralized, or hybrid), and physical domain, are the embodiments a builder can vary.

Composition Pathway

STLA Brain is unusually well-positioned to host the primitive because the consolidation already happened. The domain-controller architecture, the over-the-air update channel, and the service-oriented in-vehicle bus are the integration surfaces on which governed actuation lands. Each brand's Level-2 / Level-2+ / Level-3 configuration becomes a declared admissibility envelope, parameterized per region, expressed as a machine-checkable artifact rather than a per-variant software branch. The OTA pipeline becomes the channel through which envelope updates, a regulator clarifies an ODD, an insurer revises a liability term, a brand adds a market, are propagated to the fleet without rewriting the autonomy stack.

The composition pathway also addresses the commitment-continuity problem unique to multi-brand deployment. A Jeep, a Maserati, and a Peugeot running the same AutoDrive core but exposed through different brand experiences need a structural way to share the admissibility substrate while differentiating the authority composition. Governed actuation gives Stellantis exactly that: one primitive, many envelopes, one auditable commitment record per vehicle and per transition.

Commercial Position

Stellantis' commercial position in autonomy is not won by being the fastest to a demonstration. Mercedes already cleared that bar. It is won by being the first to certify a Level-3 envelope that scales across brands and regions without per-launch compliance reinvention. The STLA Brain thesis, one architecture across many brands, is hard to realize at the autonomy layer without an architectural substrate that lets one cleared envelope compose into many cleared deployments. Governed actuation is one such substrate. With it, a multi-brand certification story can rest on a shared, machine-checkable envelope rather than a set of parallel per-brand Level 3 programs.

Licensing Implication

The Adaptive Query licensing framework supports field-of-use licensing in passenger and light-commercial autonomous driving, with STLA Brain as the natural integration surface and AutoDrive as the application. Licensing terms can be structured to accommodate Stellantis' multi-region homologation cadence and the OTA-update model that defines software-defined-vehicle economics. The result is an AutoDrive program whose mode transitions are structurally auditable, whose envelope is regulator-legible, and whose multi-brand scaling is an architectural property rather than an organizational aspiration.

Disclosure Scope

The technology described here, including the composite admissibility evaluator and its admit, gate, defer, solicit, reject, and escalate outcomes, the graduated-actuation mode selector, the reversibility-aware commitment-point evaluator, the emergency-preemption mechanism with per-authority preemption budgets and expiration, the authority taxonomy with dynamic escalation and de-escalation, graceful degradation, and lineage-recorded actuation provenance, is disclosed in U.S. Provisional Application No. 64/049,409. This is a public, dated disclosure tied to that filing and is intended to be enabling and reasonably broad across the embodiments and variations described above, including variation in signaling medium, authority taxonomy, aggregation topology, and physical deployment domain.

References to Stellantis, STLA Brain, STLA SmartCockpit, STLA AutoDrive, the Mobile Drive joint venture, Foxconn, the named Stellantis brands and vehicles, Mercedes Drive Pilot, GM Super Cruise, Ford BlueCruise, Tesla Full Self-Driving, BYD, NIO, Xpeng, and any regulatory regime (UNECE, NHTSA/FMVSS, GB standards) or SAE level are provided solely as external market and technical context. They describe third-party products and programs as publicly reported and are not claims of, nor covered by, U.S. Provisional Application No. 64/049,409. Nothing here asserts any capability, certification, contract, incident, or schedule for any named company beyond what is publicly reported, and readers should verify current facts with primary sources.