Prompt injection in smart city and urban mobility AI

Four adversarial injection surfaces in smart city and urban mobility AI

1. Traffic and surveillance camera injection (Genetec AI Security Center, Milestone Systems XProtect AI)

Traffic and surveillance camera AI processes live video frames and captured still images from Genetec AI Security Center across more than 500,000 global installations including metropolitan traffic management centres, intersection camera networks, highway incident detection systems, and urban public safety surveillance networks that classify traffic incidents, vehicle collisions, pedestrian safety events, and criminal activity indicators from continuous camera feed images; Milestone Systems XProtect AI at more than 500,000 global installations across an 11,000+ integration partner ecosystem processing city intersection camera feeds, transit corridor surveillance images, and municipal public safety camera networks through AI-assisted incident classification, object detection, and anomalous event recognition tools; Axon AI body camera and deployed camera analytics processing law enforcement operational camera images through AI-assisted evidence classification and incident recognition tools at more than 17,000 US law enforcement agency deployments; and Esri CityEngine AI urban digital twin platforms at more than 1,000 city government customers processing city infrastructure sensor data and camera images through AI-assisted urban planning and incident detection decision support tools — extracting incident detection classifications and vehicle and pedestrian safety indicator determinations from live camera frame inputs in AI-assisted traffic management, public safety, and law enforcement decision support pipelines at metropolitan camera network scales that make individual human operator review of every AI-processed camera frame impracticable for city traffic management and law enforcement operations.

The adversarial injection surface is the live traffic and surveillance camera frame image processing pathway: Genetec AI Security Center or Milestone Systems XProtect AI traffic camera frame images submitted through AI-assisted incident classification and vehicle detection tools for AI traffic event identification and public safety alert generation. An adversarially crafted Genetec or Milestone traffic surveillance camera frame — in which pixel perturbations applied to the vehicle collision indicator display region, the pedestrian safety event visual marker, or the criminal activity pattern visual indicator in a traffic surveillance camera image cause the AI to classify a camera frame evidencing a traffic incident or public safety event as a below-threshold routine traffic image not meeting incident detection flag criteria when the actual camera frame evidences a traffic incident or public safety event meeting Genetec or Milestone AI incident detection classification criteria — can suppress an incident detection flag that would otherwise generate an emergency services dispatch recommendation, a traffic management alert, and an incident evidence archival record. In city traffic management centre environments where Genetec AI or Milestone AI processes thousands of simultaneous camera feeds without individual operator pixel-level examination of every AI-processed frame before the AI incident classification governs the traffic management operator’s emergency response workflow, adversarial suppression of incident detection indicators creates delayed emergency response and public safety consequences that generate municipal liability and law enforcement evidence integrity dimensions.

The CJIS Security Policy and Fourth Amendment consequences of adversarially suppressed incident classification in traffic and surveillance camera AI span CJIS Security Policy §5.10 criminal justice information accuracy obligations, Fourth Amendment Carpenter v. United States 138 SCt 2206 (2018) location surveillance constitutional constraints, 18 USC §2511 Electronic Communications Privacy Act interception prohibitions, and state video surveillance statute requirements. CJIS Security Policy §5.10 establishes accuracy, completeness, and timeliness requirements for criminal justice information maintained and exchanged through FBI CJIS systems — adversarial manipulation of Genetec or Milestone traffic surveillance AI that suppresses incident or vehicle detection classifications feeding into law enforcement criminal justice information records creates §5.10 accuracy obligation failures when those AI-generated classifications contribute to CJIS-connected criminal justice information. Carpenter v. United States, 138 SCt 2206 (2018), established that accessing long-term cell-site location data constitutes a Fourth Amendment search — the Court’s reasoning has been applied by lower courts to continuous surveillance camera networks that aggregate comprehensive location history data, creating Fourth Amendment warrant requirement dimensions for city-wide traffic surveillance AI deployments. 18 USC §2511 prohibits intentional interception of wire, oral, or electronic communications; adversarial manipulation of traffic surveillance AI systems that process communications infrastructure monitoring or law enforcement communications evidence images creates ECPA dimensions. Threshold: 55 for traffic surveillance AI — reflecting the CJIS §5.10 accuracy, Fourth Amendment Carpenter surveillance, 18 USC §2511 ECPA, and state video surveillance statute dimensions of adversarially manipulated incident classification.

2. Automated licence plate recognition injection (Axon AI, Genetec ALPR AI)

Automated licence plate recognition AI processes vehicle licence plate capture frames, stolen vehicle database comparison images, Amber Alert and BOLO (Be On the Lookout) hit notification displays, and state DMV vehicle registration record images from Axon AI’s ALPR platform at more than 17,000 US law enforcement agency deployments including municipal police patrol vehicles, fixed intersection ALPR readers, and law enforcement mobile ALPR units; Genetec AutoVu AI ALPR at parking enforcement operations, law enforcement deployments, and city access control installations processing licence plate capture frames through AI-assisted plate character recognition, stolen vehicle flag generation, and wanted vehicle identification pipelines; Motorola Solutions CommandCentral AI at law enforcement operations integrating ALPR data with criminal justice information systems; and TransCore AI’s EZPass-integrated electronic tolling ALPR processing more than 75% of US toll road revenue transactions through AI-assisted vehicle identification and toll account matching tools — extracting licence plate character strings, stolen vehicle flag classifications, and wanted vehicle identification indicators from ALPR capture frame inputs in AI-assisted law enforcement vehicle identification and traffic enforcement decision pipelines at patrol and intersection ALPR volumes that make individual human officer review of every AI-processed ALPR frame impracticable for law enforcement operations.

The adversarial injection surface is the ALPR vehicle licence plate capture frame image processing pathway: Axon AI or Genetec AutoVu ALPR capture frame images submitted through AI-assisted licence plate character recognition and stolen vehicle flag generation tools for AI vehicle identification and law enforcement alert generation. An adversarially crafted ALPR frame — in which pixel perturbations applied to the licence plate character display region, the plate state jurisdictional indicator visual marker, or the plate condition and angle display in an ALPR capture frame image cause the AI to misclassify a plate string matching an active stolen vehicle entry or Amber Alert associated vehicle registration as a non-matching plate character string that does not trigger the stolen vehicle flag or Amber Alert hit notification when the actual ALPR frame captures a plate that matches the stolen vehicle or Amber Alert registration — can suppress a stolen vehicle flag or Amber Alert notification that would otherwise generate a law enforcement pursuit initiation, an Amber Alert response dispatch, and a criminal justice NCIC hit record. In patrol ALPR environments where Axon AI or Genetec AutoVu AI processes hundreds of ALPR captures per officer shift without individual officer pixel-level examination of every AI-processed ALPR frame before the AI plate classification governs the officer’s stolen vehicle response workflow, adversarial suppression of stolen vehicle flag indicators creates law enforcement response delay consequences with criminal justice information accuracy and civil liability dimensions.

The CJIS, FCRA, and Driver’s Privacy Protection Act consequences of adversarially manipulated plate character classification in ALPR AI span CJIS Security Policy NCIC accuracy obligations, FCRA 15 USC §1681e(b) maximum possible accuracy requirements, Driver’s Privacy Protection Act 18 USC §§2721–2725 DMV data use limitations, and state ALPR data retention statute requirements. CJIS NCIC accuracy obligations require that criminal justice information submitted to and obtained from NCIC be accurate, complete, and based on reliable information sources — adversarial manipulation of Axon or Genetec ALPR AI that suppresses plate matches against NCIC stolen vehicle entries creates CJIS NCIC accuracy obligation failures when the AI-generated plate classification contributes to or omits NCIC-connected criminal justice information records. FCRA 15 USC §1681e(b) requires consumer reporting agencies to follow reasonable procedures to assure maximum possible accuracy of information concerning the individual reported — where ALPR-derived vehicle and owner information enters consumer reporting workflows through background screening or credit decisioning systems, adversarial ALPR AI manipulation creates §1681e(b) accuracy obligation dimensions. The Driver’s Privacy Protection Act, 18 USC §§2721–2725, restricts disclosure of DMV-held personal information about drivers; adversarial manipulation of ALPR AI that generates inaccurate plate-to-owner matching classifications affecting DMV data access creates DPPA data use limitation dimensions for law enforcement and commercial ALPR operators. Threshold: 60 for ALPR AI — reflecting CJIS NCIC accuracy, FCRA §1681e(b) maximum accuracy, DPPA DMV data use, and state ALPR retention statute dimensions.

3. Parking enforcement camera injection (ParkMobile AI, TransCore AI)

Parking enforcement camera AI processes vehicle licence plate capture photographs, parking meter display images, accessible parking space occupancy photographs, and street-cleaning zone violation images from ParkMobile AI at more than 4,000 cities with 30 million+ registered users processing parking enforcement camera photographs through AI-assisted plate recognition and parking violation classification tools that municipal parking enforcement officers and automated enforcement vehicles use for citation initiation at city parking enforcement volumes; Passport Labs AI at municipal parking operations processing parking enforcement camera images through AI-assisted parking violation detection and citation workflow initiation tools; TransCore AI electronic tolling ALPR at tolling operations integrated with city parking enforcement and access control systems; and Conduent Transportation AI at parking management operations including transit parking facilities and municipal parking authority enforcement systems processing parking enforcement camera images through AI-assisted violation classification and citation generation pipelines — extracting parking violation classification determinations and citation initiation indicators from enforcement camera photograph inputs in AI-assisted municipal parking enforcement and revenue management pipelines at urban enforcement volumes that make individual human officer examination of every AI-processed parking enforcement photograph impracticable for city parking operations.

The adversarial injection surface is the parking enforcement camera photograph image processing pathway: ParkMobile AI or TransCore AI parking enforcement camera photograph images submitted through AI-assisted parking violation classification and citation initiation tools for AI enforcement determination and citation record generation. An adversarially crafted ParkMobile or Passport Labs parking enforcement camera photograph — in which pixel perturbations applied to the parking meter payment display region, the licence plate character indicator visual marker, or the time-limited zone eligibility display in a parking enforcement camera photograph cause the AI to classify a vehicle occupying a parking space in violation of municipal parking time limits, meter payment requirements, or accessible space reservation requirements as a compliant vehicle not meeting parking citation threshold criteria when the actual photograph documents a parking violation meeting municipal ordinance citation issuance criteria — can suppress a violation detection that would otherwise generate a parking citation initiation, a parking enforcement officer citation workflow input, and a municipal revenue record. In automated parking enforcement vehicle environments where ParkMobile AI processes hundreds of parking enforcement photographs per enforcement vehicle tour without individual officer pixel-level examination of every AI-processed photograph before the AI violation classification governs the citation initiation workflow, adversarial suppression of violation indicators creates municipal parking ordinance enforcement revenue loss and accessible parking space enforcement failure dimensions.

The municipal ordinance enforcement, ADA, and due process consequences of adversarially suppressed violation classification in parking enforcement camera AI span municipal parking ordinance enforcement revenue integrity, ADA 42 USC §12101 accessible parking space enforcement obligations, 14th Amendment due process notice requirements for parking citation issuance, and state parking enforcement statute administrative compliance dimensions. Municipal parking ordinances establish parking time limits, meter payment requirements, accessible parking space reservation requirements, and street-cleaning zone restrictions that generate citation revenue and enforce ADA-required accessible parking space availability; adversarial manipulation of ParkMobile or Passport Labs parking enforcement AI that suppresses violation detections reduces citation issuance rates and erodes municipal parking ordinance enforcement revenue and accessible parking space compliance enforcement. ADA 42 USC §12101 establishes comprehensive disabled individuals’ rights including accessible parking space access rights enforced through municipal parking citation issuance for accessible space violations; adversarial manipulation of parking enforcement camera AI that suppresses accessible space occupancy violation detections creates ADA accessible space enforcement failure dimensions affecting disabled individuals’ parking access rights. The 14th Amendment due process clause requires that parking citations be issued only when supported by adequate evidence of a parking violation; adversarially generated ParkMobile AI citation records that falsely document non-violating vehicles as violating vehicles — through adversarial manipulation that creates false positive violation indicators rather than suppressing real ones — create due process notice challenges for citation contest proceedings. Threshold: 65 for parking enforcement AI — reflecting municipal ordinance enforcement revenue, ADA §12101 accessible space enforcement, 14th Amendment due process citation notice, and state parking enforcement statute dimensions.

4. Transit fare evasion detection injection (Cubic Transportation Systems AI, Conduent Transportation AI)

Transit fare evasion detection AI processes fare gate camera images, turnstile passage photographs, transit platform surveillance images, and transit payment validation displays from Cubic Transportation Systems AI at Chicago Ventra, London Oyster, New York MTA, and Sydney Opal transit fare systems processing fare gate camera images and passenger fare payment validation images through AI-assisted fare evasion detection and fare gate access control tools; Conduent Transportation AI processing more than 50 million daily transit rider interactions across US and international transit authority operations including bus fare validation, rail fare gate, and transit platform surveillance AI systems; and HERE Technologies AI mobility mapping tools at transit authority operations integrating fare gate data with transit network mobility intelligence platforms — extracting fare evasion detection classifications and fare payment compliance indicator determinations from fare gate camera frame inputs in AI-assisted transit fare enforcement and revenue integrity decision pipelines at transit system throughput volumes that make individual human transit officer review of every AI-processed fare gate image impracticable for high-throughput urban transit systems. Cubic Transportation Systems AI fare gate platforms process millions of daily passenger fare validation events at transit systems where fare evasion detection AI governs transit authority response protocols, fare enforcement officer dispatch, and cumulative fare revenue calculations that feed into FTA §5307 Urbanized Area Formula Grants programme compliance reporting.

The adversarial injection surface is the transit fare gate camera image processing pathway: Cubic Transportation Systems AI or Conduent Transportation AI fare gate camera frame images submitted through AI-assisted fare evasion detection and fare payment compliance classification tools for AI fare enforcement determination and transit revenue integrity management. An adversarially crafted Cubic or Conduent fare gate camera image — in which pixel perturbations applied to the fare payment validation display region, the transit pass or ticket media visual indicator, or the physical fare gate passage behaviour indicator in a fare gate camera image cause the AI to classify a passenger evading fare payment by tailgating, jumping a turnstile, or bypassing a fare gate as a compliant fare-paying passenger not meeting fare evasion detection threshold criteria when the actual camera image documents a fare evasion event meeting Cubic or Conduent AI fare evasion detection criteria — can suppress a fare evasion detection that would otherwise generate a fare enforcement officer dispatch notification, a fare gate access control record, and a transit revenue integrity audit entry. In high-throughput transit system environments where Cubic AI or Conduent AI processes thousands of simultaneous fare gate camera images per station without individual transit officer pixel-level examination of every AI-processed gate image before the AI fare evasion classification governs the fare enforcement response workflow, adversarial suppression of fare evasion indicators creates systematic transit revenue shortfall consequences with FTA grant compliance and ADA equal access dimensions.

The FTA, ADA, and transit revenue integrity consequences of adversarially suppressed fare evasion classification in transit fare AI span FTA 49 USC §5307 Urbanized Area Formula Grants fare revenue compliance obligations, ADA 42 USC §12101 equal transit access requirements, transit authority fare revenue integrity obligations, and state transit authority statute administrative compliance dimensions. FTA 49 USC §5307 Urbanized Area Formula Grants provide federal formula funding to urban transit authorities for public transportation operations and capital projects; FTA grant programme requirements include fare revenue compliance obligations and financial management standards that transit authorities must satisfy as conditions of federal grant receipt — adversarial manipulation of Cubic or Conduent transit fare evasion AI that systematically suppresses fare evasion detections reduces transit authority fare revenue, creates underreported fare evasion rate metrics, and creates FTA financial management and fare revenue compliance reporting dimensions that affect grant programme eligibility. ADA 42 USC §12101 requires that public transportation systems be accessible to disabled individuals and that transit services not discriminate against disabled passengers; adversarial manipulation of transit fare evasion AI that creates discriminatory fare enforcement patterns — for example, adversarial inputs targeting fare evasion detection at accessible entrances and reduced-fare validation points — creates ADA equal transit access enforcement dimensions for transit authorities that receive FTA funding. Transit authorities bear contractual and statutory obligations to maintain fare revenue integrity and accurate fare collection records for FTA grant programme financial reporting; adversarially corrupted Cubic or Conduent fare evasion AI that generates systematically inaccurate fare evasion detection records creates financial reporting accuracy obligations and potential grant compliance exposure. Threshold: 65 for transit fare evasion AI — reflecting FTA 49 USC §5307 grant compliance, ADA §12101 equal transit access, transit revenue integrity, and state transit authority statute dimensions.

Integration: smart city and urban mobility AI image ingestion with Glyphward pre-scan

Smart city and urban mobility AI image ingestion flows from Genetec AI Security Center and Milestone Systems XProtect AI traffic surveillance camera frame channels, Axon AI and Genetec AutoVu ALPR capture frame interfaces, ParkMobile AI and TransCore AI parking enforcement camera photograph platforms, and Cubic Transportation Systems AI and Conduent Transportation AI transit fare gate camera image processing systems into traffic incident detection AI, stolen vehicle flag generation AI, parking violation classification AI, and fare evasion detection AI pipelines. Insert Glyphward’s pre-scan at the ingestion boundary before AI-generated output is committed to traffic incident alerts, ALPR stolen vehicle flags, parking enforcement citation initiations, or transit fare evasion enforcement records:

import asyncio
import base64
import hashlib
import os
import uuid
from enum import Enum
from pathlib import Path

import httpx

GLYPHWARD_API_KEY = os.environ["GLYPHWARD_API_KEY"]
GLYPHWARD_SCAN_URL = "https://glyphward.com/v1/scan"

# Smart city & urban mobility AI — CJIS §5.10 criminal justice information accuracy;
# 4th Amendment Carpenter v. United States 138 SCt 2206 (2018);
# 18 USC §2511 ECPA; FCRA 15 USC §1681e(b); DPPA 18 USC §§2721-2725;
# FTA 49 USC §5307 Urbanized Area Formula Grants; ADA 42 USC §12101.
THRESHOLD_TRAFFIC_SURVEILLANCE_AI   = 55  # Genetec/Milestone; CJIS §5.10; 4th Amend; ECPA
THRESHOLD_LICENCE_PLATE_RECOGNITION_AI = 60  # Axon/Genetec ALPR; CJIS NCIC; FCRA; DPPA
THRESHOLD_PARKING_ENFORCEMENT_AI    = 65  # ParkMobile/TransCore; ADA §12101; 14th Amend
THRESHOLD_TRANSIT_FARE_EVASION_AI   = 65  # Cubic/Conduent; FTA §5307; ADA §12101


class SmartCityUrbanMobilityAIContext(str, Enum):
    TRAFFIC_SURVEILLANCE_AI        = "traffic_surveillance_ai"         # Genetec, Milestone
    LICENCE_PLATE_RECOGNITION_AI   = "licence_plate_recognition_ai"    # Axon, Genetec AutoVu
    PARKING_ENFORCEMENT_AI         = "parking_enforcement_ai"          # ParkMobile, TransCore
    TRANSIT_FARE_EVASION_AI        = "transit_fare_evasion_ai"         # Cubic, Conduent


def threshold_for(context: SmartCityUrbanMobilityAIContext) -> int:
    mapping = {
        SmartCityUrbanMobilityAIContext.TRAFFIC_SURVEILLANCE_AI:        THRESHOLD_TRAFFIC_SURVEILLANCE_AI,
        SmartCityUrbanMobilityAIContext.LICENCE_PLATE_RECOGNITION_AI:   THRESHOLD_LICENCE_PLATE_RECOGNITION_AI,
        SmartCityUrbanMobilityAIContext.PARKING_ENFORCEMENT_AI:         THRESHOLD_PARKING_ENFORCEMENT_AI,
        SmartCityUrbanMobilityAIContext.TRANSIT_FARE_EVASION_AI:        THRESHOLD_TRANSIT_FARE_EVASION_AI,
    }
    return mapping[context]


async def scan_smart_city_urban_mobility_ai_image(
    image_path: str | Path,
    context: SmartCityUrbanMobilityAIContext,
    city_agency_id_hash: str,       # SHA-256 of city agency or transit authority identifier
    asset_or_vehicle_ref: str,      # e.g. "CAM-INT-2026-44821", "PLATE-CA-2026-88841"
    monitoring_session_id: str,     # camera feed batch, ALPR session, enforcement tour ID
    client: httpx.AsyncClient,
) -> dict:
    """
    Scan a smart city or urban mobility AI image for adversarial injection payloads
    before forwarding to traffic surveillance incident detection, ALPR stolen vehicle
    flag generation, parking violation classification, or transit fare evasion detection
    AI systems.

    Raises AdversarialSmartCityUrbanMobilityAIImageError if score meets threshold:
      - TRAFFIC_SURVEILLANCE_AI:       threshold 55; CJIS §5.10; 4th Amend Carpenter; ECPA
      - LICENCE_PLATE_RECOGNITION_AI:  threshold 60; CJIS NCIC; FCRA §1681e(b); DPPA
      - PARKING_ENFORCEMENT_AI:        threshold 65; ADA §12101; 14th Amend due process
      - TRANSIT_FARE_EVASION_AI:       threshold 65; FTA §5307; ADA §12101; revenue integrity
    """
    image_bytes     = Path(image_path).read_bytes()
    image_b64       = base64.b64encode(image_bytes).decode()
    image_sha256    = hashlib.sha256(image_bytes).hexdigest()
    client_scan_id  = str(uuid.uuid4())
    threshold       = threshold_for(context)

    resp = await client.post(
        GLYPHWARD_SCAN_URL,
        headers={"Authorization": f"Bearer {GLYPHWARD_API_KEY}"},
        json={
            "image": image_b64,
            "source": context.value,
            "metadata": {
                "smart_city_urban_mobility_context": context.value,
                "city_agency_id_hash":               city_agency_id_hash,
                "asset_or_vehicle_ref":              asset_or_vehicle_ref,
                "monitoring_session_id":             monitoring_session_id,
                "client_scan_id":                    client_scan_id,
                "image_sha256":                      image_sha256,
            },
        },
        timeout=8.0,
    )
    resp.raise_for_status()
    result = resp.json()

    audit_record = {
        "city_agency_id_hash":              city_agency_id_hash,
        "asset_or_vehicle_ref":             asset_or_vehicle_ref,
        "monitoring_session_id":            monitoring_session_id,
        "smart_city_urban_mobility_context": context.value,
        "scan_id":                          result["scan_id"],
        "client_scan_id":                   client_scan_id,
        "image_sha256":                     image_sha256,
        "score":                            result["score"],
        "flagged_region":                   result.get("flagged_region"),
        "threshold":                        threshold,
        "action":                           "blocked" if result["score"] >= threshold else "allowed",
    }
    await write_urban_mobility_audit_record(audit_record)

    if result["score"] >= threshold:
        raise AdversarialSmartCityUrbanMobilityAIImageError(
            f"Smart city AI image blocked [{context.value}]: "
            f"scan_id={result['scan_id']} score={result['score']} "
            f"agency={city_agency_id_hash} ref={asset_or_vehicle_ref}"
        )
    return result


async def write_urban_mobility_audit_record(record: dict) -> None:
    """Persist audit record to urban mobility AI compliance documentation store (stub)."""
    import json, sys
    print(json.dumps(record), file=sys.stderr)


class AdversarialSmartCityUrbanMobilityAIImageError(Exception):
    """Raised when a smart city or urban mobility AI image exceeds the adversarial injection threshold."""
    pass

Call scan_smart_city_urban_mobility_ai_image() with SmartCityUrbanMobilityAIContext.TRAFFIC_SURVEILLANCE_AI before forwarding Genetec AI Security Center or Milestone Systems XProtect AI traffic camera frames to incident detection and vehicle classification AI — with asset_or_vehicle_ref linking the Glyphward scan to the camera asset record for CJIS §5.10 accuracy, Fourth Amendment Carpenter surveillance, and 18 USC §2511 ECPA compliance documentation. Call with SmartCityUrbanMobilityAIContext.LICENCE_PLATE_RECOGNITION_AI for Axon AI or Genetec AutoVu ALPR capture frame images before AI plate character recognition and stolen vehicle flag generation, with city_agency_id_hash for CJIS NCIC accuracy, FCRA §1681e(b) maximum possible accuracy, and DPPA §§2721–2725 DMV data use limitation audit trail documentation. Call with SmartCityUrbanMobilityAIContext.PARKING_ENFORCEMENT_AI for ParkMobile AI or Passport Labs AI parking enforcement camera photographs before violation classification and citation initiation, with monitoring_session_id as the enforcement vehicle tour identifier for ADA §12101 accessible space enforcement and 14th Amendment due process citation notice compliance documentation. Call with SmartCityUrbanMobilityAIContext.TRANSIT_FARE_EVASION_AI for Cubic Transportation Systems AI or Conduent Transportation AI fare gate camera images before fare evasion detection classification, with city_agency_id_hash for FTA §5307 Urbanized Area Formula Grants compliance, ADA §12101 equal transit access, and transit revenue integrity audit trail. Get early access

Coverage matrix

Frequently asked questions

Further reading

• CCTV and physical security AI prompt injection — related attack surface covering adversarial injection in physical security camera AI systems with access control, intrusion detection, and CJIS criminal justice information accuracy dimensions applicable to Genetec and Milestone security deployment contexts.
• Government and public sector AI prompt injection — broader regulatory framework covering 42 USC §1983 civil rights, Fourth Amendment, Brady disclosure, NVRA, and HAVA obligations applicable to city and transit authority AI deployments integrated with law enforcement and voter services operations.
• Autonomous vehicle fleet safety AI prompt injection — related adversarial attack surface covering AI injection in autonomous and connected vehicle systems with HERE Technologies mobility mapping, vehicle-to-infrastructure communication, and NHTSA safety regulation dimensions.
• Free tier — 10 scans/day, no card required — start scanning smart city and urban mobility AI images at development volumes before committing to a production plan.