AutoX — agentic threat model

Not certain from the listing — The listing mentions 'AI driving systems' but does not specify the exact foundation models or neural network architectures used. Threats include adversarial physical attacks (e.g., adversarial stickers on road signs) and model evasion.

Not certain from the listing — The listing details sensor modalities (cameras, LiDAR, radar, IMUs, GPS) but does not describe the data operations, training pipelines, or local data storage. Threats include sensor data spoofing and training data poisoning.

Not certain from the listing — The proprietary orchestration and decision-making framework for Level 4 autonomy is not detailed. Threats include exploitation of the path-planning logic or hijacking of the vehicle control commands.

Not certain from the listing — The physical vehicle hosting environment and onboard edge-compute infrastructure are not specified. Threats include physical access to the vehicle's internal networks (e.g., CAN bus) and remote exploitation of wireless interfaces.

Not certain from the listing — Real-time safety guardrails, logging, and observability systems are not detailed in the public listing. Threats include silent perception failures, sensor drift, and lack of out-of-distribution detection in novel driving environments.

Not certain from the listing — The listing does not mention specific automotive cybersecurity standards (such as ISO/SAE 21434) or functional safety standards (ISO 26262). Threats include regulatory non-compliance and lack of standardized AI safety audits.

Not certain from the listing — While a 'robotaxi service' implies a centralized dispatch or fleet management system, specific multi-agent coordination or V2X protocols are not detailed. Threats include fleet-wide compromise via a single compromised dispatch agent.