phy·lax /ˈfaɪlæks/
noun  ·  Ancient Greek  φύλαξ
1. a guardian; a watchman; a sentinel.
"a phylax stood watch at the city walls"
AI Security · South Africa & Beyond

Stop false alarms.
Catch real threats.

98% of security alerts are false. Phylax Vision uses on-device AI to identify actual persons, vehicles, and intrusions — while ignoring everything else. No cloud. No subscription fees. No missed incidents.

Request Investor Deck ↗ View Platform →
<5%
false positive rate
<50ms
detection latency — on device
0
cloud dependency for AI
Phylax Vision Viewer v1.0.0 — Live Intelligence
Phylax Vision Viewer — live camera feeds with AI detection

Live multi-camera feed — actual deployment screenshot

The Problem

Security cameras everywhere.
Actual security — almost nowhere.

The global security industry has spent decades installing motion sensors, PIR detectors, and cameras — yet property crime rates remain stubbornly high. The reason is not a lack of cameras. It is a fundamental flaw in how those cameras decide when something is wrong.

98%[1]
of all security camera alerts are false alarms — leaves, shadows, headlights, spiders
$1.8B[1]
wasted annually in the US alone — police responding to 36 million false alarm calls
R400+[2]
per false armed-response call-out in South Africa — a direct cost customers pay every time
<5%[3]
false positive rate with AI object detection — vs 80–98% with legacy motion sensors
Legacy motion detection — how it fails
  • ✕Triggers on any pixel change or heat source — a passing car, a cloud shadow, rain on the lens
  • ✕Cannot distinguish between a person and a tree blowing in the wind
  • ✕Alarm fatigue sets in — owners disable alerts or stop responding to them entirely
  • ✕No context — cannot classify what was seen, only that something changed
  • ✕Criminals learn the weak zones and timing patterns of static motion grids
Phylax Vision AI — how it solves this
  • ✓Every camera frame is analysed for specific objects — person, car, truck, motorcycle
  • ✓Stationary person at the gate = alert. Blowing tree = silence. The AI knows the difference
  • ✓Processing runs entirely on the device — sub-50 ms response, no cloud, no latency
  • ✓Three-mode response system — Disarmed / Notify / Armed — no cry-wolf fatigue
  • ✓AI is continuously retrained as criminal behaviour adapts — not a static product
⚡
Why on-device matters — edge vs cloud
Cloud-based AI adds 500–700 ms of latency per detection[4] — enough time for a person to cross your property line before an alert is raised. Phylax Vision runs inference directly on the phylax device, reducing detection latency to under 50 ms. There is no cloud subscription, no third-party data exposure, and no dependency on an internet connection being up during an incident.
Live Deployment

This is not a demo.
It is running right now.

Every screenshot below is taken from a live phylax.vision deployment — five cameras, armed, with over 93,000 AI detections recorded. No mockups. No staged data.

Web Dashboard Live system — actual running deployment
93,228 total detections 47h uptime 5 cameras · ARMED
Dashboard — system status, detection counts, inference timing, camera health
System Dashboard — AI status, inference metrics, 24h detection activity
AI Detections — thumbnail grid of all detection events
AI Detections — searchable thumbnail grid with camera and class filters
Setup — detection classes, confidence threshold, camera config
Setup — detection classes, confidence tuning, camera management
Detection Workflows — automation triggers and execution log
Detection Workflows — event-driven automation with live execution log
Mobile App Actual screenshots — phylax.vision.app v1.0.0 running on Android
Home — live feeds and armed status
Home — live feeds & status
Camera detail and detection grid
Camera detail & detections
Detection notifications list
Detection notifications
Data usage monitoring
Data usage monitoring
Why Not the Incumbents?

Existing solutions leave
the gap wide open.

The security camera market is not short of products. It is short of products that actually eliminate false alarms, run AI on-device, and give a security company a ready-made platform to resell. Here is why the alternatives fall short.

Solution AI Detection On-Device / No Cloud Works With Existing Cameras B2B Reseller-Ready
phylax.vision ✓ Object recognition — on device ✓ Fully on-device, no cloud ✓ Any existing IP camera ✓ API + white-label ready
Basic / legacy CCTV + NVR[8] ✕ Pixel-change motion only — no object classification ✓ On-site ✓ Any camera ✕ No software platform
Enterprise VMS — Milestone, Genetec[9] ✓ AI analytics ~ Server or edge — enterprise licensing required ~ Camera-dependent per tier ✕ Enterprise pricing only — not SME/residential
AI cameras — Hikvision AcuSense / Dahua[10] ~ On-camera AI — own cameras needed for full capability ✓ On-camera processing ~ Limited channels with third-party cameras ✕ Proprietary closed ecosystem
Consumer AI — Ring, Arlo[11] ~ Basic detection — paid cloud subscription required for AI ✕ Cloud subscription mandatory for AI features ✕ Own branded hardware only ✕ Consumer retail only
The gap: No competitor offers AI object detection that runs entirely on-device, works with a customer's existing IP cameras, and ships as a resellable platform for security companies — at a price point accessible to residential and SME markets.
The Platform

Four integrated products.
One complete system.

Every layer of the stack is owned by phylax.vision — from the AI inference engine on the edge device to the mobile app in your pocket.

⚙️
phylax device
Edge · Cluster · On-Premise
ARM64 x86_64 Edge Cluster
Edge Device Agent
phylax.vision.device
Intelligent on-premise surveillance agent. Runs on edge hardware with a dedicated AI accelerator, or on server-grade hardware for multi-camera cluster deployments. Handles camera ingestion, real-time AI inference, event storage, and secure cloud reporting.
Hardware AI acceleration — edge & cluster
GPU-accelerated cluster node support
Wide IP camera compatibility
Detection workflows with cloud integration
Protected model delivery — device-locked
OTA binary + model self-updates
phylax.vision web dashboard
Web Dashboard
phylax.vision.dashboard
Browser-based control interface served directly from the phylax device. Live camera streams, AI detection feed, configuration, detection workflows, and multi-user access management — no app install required.
Live Intelligence — up to 6 camera streams
AI Detections tab — thumbnail grid + lightbox
Detection Workflows — configurable automation
Three security modes: Disarmed · Notify · Armed
Role-based user access (admin, installer, viewer)
Real-time device health & stats
phylax.vision desktop viewer
Desktop Viewer
phylax.vision.viewer
Native desktop application for professional on-site monitoring. Connects directly to phylax devices on the local network for low-latency live streams, a full detection timeline, and audible alerts when detections occur.
Native Linux, Windows, macOS
Local LAN connection — ultra-low latency
Audible sound alerts on detection
Detection timeline + thumbnail strip
Smooth browsing of thousands of detections
Per-camera activity graphs
phylax.vision mobile app — home screen
Mobile App
phylax.vision.app
Mobile application for Android and iOS. Connects through the phylax.vision cloud to provide secure remote access to any registered phylax device — anywhere in the world, without port forwarding.
Android & iOS native
Push notifications on detection
Biometric authentication
QR code device claiming
Multi-device & group sharing
24h detection activity graphs
Live camera stream per device
WhatsApp group alert linking
Data usage monitoring
Architecture

Built for the edge.
Ready for the cloud.

AI inference runs entirely on-device — no footage ever leaves your property. The phylax.vision.server provides secure provisioning, model updates, and connectivity. Cloud integration is available for enterprise workflows.

🧠
Hardware-Accelerated AI Inference
Dedicated AI accelerator hardware delivers real-time inference at ultra-low power on edge nodes. GPU-accelerated cluster deployments scale to hundreds of cameras without cloud dependency.
🧬
In-House Trained AI — Continuously Improving
Phylax Vision trains its own AI detection models from scratch — no dependency on third-party libraries or licensed weights. Accuracy improves with every training cycle, and models can be retrained to counter emerging threats as criminal behaviour evolves. Our AI adapts; it is never a static product.
🔐
Encrypted Model Delivery
AI models are cryptographically protected and keys are derived at runtime — the raw model never touches disk. Each device is uniquely bound to its licensed model, making redistribution or cloning technically impossible.
🔄
Automatic OTA Updates
Both the device software and AI model files update automatically from a signed manifest. Background downloads, atomic swaps, and one-step rollback ensure zero-downtime updates across all deployed devices.
🌐
Secure Mesh — No Port Forwarding
Each device automatically joins a proprietary secure mesh network on provisioning. The mobile app connects through the mesh — encrypted, NAT-traversing — with no port forwarding, no static IP, and no complex network configuration required.
☁️
Cloud Integration
Detection workflows on the device support configurable automation steps executed as shell commands or HTTP calls — enabling integration with MQTT brokers, cloud platforms (AWS, Azure, Google Cloud), webhooks, and custom enterprise systems. Remotes can be defined with API key authentication for secure multi-site cloud reporting.
System Architecture
Mobile & Desktop Clients
Mobile App (Android / iOS) Desktop Viewer (Win / Mac / Linux)
↕ HTTPS + JWT
phylax.vision.server
Auth & Sessions Device Provisioning Model Crypto OTA Updates Push Notifications WhatsApp Groups Invite System Group RBAC DNS Automation
↕
Database
Users · Devices · Detections · Groups · Logs
Infrastructure
Cache Secure Mesh VPN Push Services
↕ Signed Manifest · Protected Model Blob
phylax.vision.device — Edge / Cluster
AI Accelerator GPU Cluster Camera Ingest Live Streaming Event Storage Workflows Web Dashboard Auto Discovery
↕ IP Camera Protocol
IP Cameras
Front Door Driveway Back Yard + N cameras
AI Detection Technology

Not motion detection.
Intelligent object recognition.

Traditional security systems trigger on any movement — a leaf blowing, headlights sweeping a wall, a spider on the lens. Phylax Vision works fundamentally differently. Every camera frame is analysed by our AI engine, which identifies and classifies specific objects — persons, vehicles, motorcycles — regardless of movement. A stationary person standing at your gate triggers a detection. A tree blowing in the wind does not.

AI detects a person in the back yard at night — infrared night vision with bounding box and confidence score
person · 88% Person identified in complete darkness — infrared night vision — 23:15
AI detects a person on the patio — bounding box with confidence score
person · 81% Person detected at the entertainment area — bounding box and confidence overlay — 16:11
🌙
Night Vision & Infrared
Fully operational in complete darkness. Compatible with infrared (IR) cameras — the AI analyses the greyscale IR image and detects persons and vehicles with the same accuracy as daylight.
🌩️
All-Weather Operation
Trained and tested under rain, fog, overcast skies, harsh shadows, and strong direct sunlight. Object recognition is robust to the difficult lighting conditions common in Southern Africa.
🔥
Thermal / FLIR Compatible
Supports thermal imaging cameras (FLIR and similar). Heat-signature imagery enables detection in conditions where standard IR cameras struggle — total darkness, heavy smoke, or dense fog.
🎯
Per-detection confidence scores & bounding boxes
Every detection includes a confidence percentage and precise bounding box coordinates. Adjustable confidence threshold lets you tune sensitivity — raise it to suppress low-confidence detections, lower it to catch partially occluded or distant subjects. Detections are recorded with a full-frame snapshot and timestamp for audit and evidence purposes.
person car truck bus motorcycle bicycle
AI Intelligence

Four model tiers.
Matched to your hardware.

Phylax Vision ships with separate AI model tiers for edge devices and cluster nodes. The server manages versioned model delivery and automatic upgrades — devices always run the latest approved model without manual intervention.

Scroll to explore all model tiers

Edge · Accelerated
Large
Optimised for edge hardware — balanced accuracy and real-time speed
Inference SpeedFast
Detection AccuracyHigh
Power EfficiencyExcellent
Classes: person · car · bus · truck · bicycle · motorcycle
Device-locked · OTA versioned
Edge · Accelerated
Extra LargeXL
Maximum accuracy — for critical deployments where detection rate matters most
Inference SpeedModerate
Detection AccuracyMaximum
Power EfficiencyGood
Classes: person · car · bus · truck · bicycle · motorcycle + extended
Device-locked · OTA versioned
Server Node · x86_64
Edge Server
For server-grade deployments — CPU inference, no dedicated accelerator required
CompatibilityUniversal
Detection AccuracyHigh
Deploy CostLow
Runs on any standard x86_64 server · No GPU required
Device-locked · OTA versioned via server manifest
GPU Cluster · Multi-Site
ClusterGPU
GPU-accelerated cluster nodes — for high-density multi-camera enterprise deployments
ThroughputMaximum
Detection AccuracyMaximum
ScalabilityEnterprise
GPU-accelerated inference · Multi-camera cluster · Horizontal scaling
Device-locked · OTA versioned · Enterprise grade
🔄
Automatic Model Versioning
The server independently tracks versions for each model tier — edge accelerated (Large & XL) and cluster (server node & GPU). Each device polls hourly, downloads updates to a staging area, and performs an atomic swap. Failed updates automatically roll back to the previous model — zero-downtime guaranteed.
Capabilities

Everything built in.
Nothing bolted on.

Scroll to browse all capabilities

🎯
Detection Workflows
Visual workflow editor triggers customisable shell commands or HTTP calls on any detection event — per camera, per class, with configurable confidence thresholds, cooldown periods, and retry logic.
👥
Multi-User & Groups
Full role-based access: admin, installer, and viewer roles. Group any set of devices together. Invite users by email with permission-scoped tokens. Biometric auth on mobile.
📱
Push Notifications
Firebase Cloud Messaging delivers real-time detection alerts to the mobile app. Class-filtered push — only notify on the events that matter: person, car, vehicle intrusion.
💬
WhatsApp Group Notifications
Link any phylax group to one or more WhatsApp groups. Every detection fires an alert directly into the chat — with the AI-annotated camera snapshot, camera name, detected class, and confidence score. No app install needed on the recipient side.
🔒
Privacy First — True Edge AI
AI runs on-device. Camera footage is never sent to any cloud for processing. POPIA compliant by design. The cloud server handles metadata only — no video, no frames.
📡
Zero-Config Remote Access
Secure mesh VPN provisioned automatically on device registration. No port forwarding, no static IP, no NAT punch-through complexity. Works behind CGNAT and 4G/LTE connections.
📊
Complete Observability
24-hour rolling detection sparklines, CPU/RAM/temperature gauges, per-camera FPS/uptime metrics, full structured server logs, and data usage tracking across all stream categories.
🔔
Three-Mode Security System
Disarmed — cameras stream, no alerts. Notify — detections logged and pushed to mobile, no alarm. Armed — full alert mode with push notifications, audible alerts, and workflow triggers. Switch modes instantly from any device.
🧬
Proprietary AI Models — In-House Trained
Phylax Vision trains its own AI detection models — we are not dependent on third-party libraries. This means accuracy improves continuously with every new training run, and we can retrain to counter new threat patterns as criminals adapt. Our AI is never a black box we cannot improve.
⏰
Scheduled Arming
Per-camera arming schedules — day of week and time windows. Set your front door camera to armed 24/7 and pool camera to notify only during school hours.
🏗️
Multi-Site Architecture
Register unlimited devices under one phylax.vision.server instance. DNS automation assigns each device a unique subdomain. Scale from 1 to 1,000 sites without infrastructure changes.
🔍
Automatic Device Discovery
Local network discovery lets the desktop viewer and dashboard find phylax devices automatically without manual IP configuration — zero touch onboarding for installers.
📅
Detection Timeline
Every detection is stored with its full-frame snapshot, timestamp, detected class, and confidence score. The timeline is searchable and filterable by camera and class — giving you a chronological audit trail of every event on every camera, with one-click lightbox review.
Integration & Automation

A detection platform,
not just a camera system.

Phylax Vision is built to be the intelligent core of a larger security ecosystem — not a closed box. It connects outward to professional monitoring centres and downward to physical hardware on-site.

🏢
Security Monitoring Station Integration

Phylax Vision is designed to feed directly into professional central monitoring stations and armed response control rooms. When a detection triggers in Armed mode, the event — with timestamp, camera, classification, and snapshot — can be relayed instantly to a monitoring operator without them ever touching the camera system.

→ Webhook / REST API — push real-time detection events to any monitoring platform, CRM, or dispatch system
→ WhatsApp & SMS routing — detection alerts with snapshot images delivered directly to armed response operators
→ Email & cloud service triggers — notifications to any SMTP, Slack, or third-party operations tool
→ Branded white-label model — security companies can deploy phylax devices under their own brand, powered by our AI
→ Multi-site operator view — a single server instance can monitor hundreds of client sites, each with their own device
For security companies: Add AI detection capability to your existing monitoring offering without replacing client camera hardware. One phylax device per site, integrated into your control room via API.
⚙️
On-Site Hardware Automation

The phylax device exposes hardware interface ports that allow it to directly control or monitor physical equipment on-site — transforming a detection event into an immediate physical response, without any additional controller or hub.

→ Floodlight / deterrent lights — switch on automatically the instant a person is detected
→ External siren & strobe — trigger a wired alarm siren or beacon directly from the device
→ Gate & door control — integrate with electric gate motors or access control lock releases
→ Electric fence integration — trigger or energiser-link via relay output
All triggered by AI detection: A verified person detection at 23:00 can simultaneously alert your phone, notify your armed response company, switch on all exterior lights, and trigger a siren — in under one second.
💬
WhatsApp Group Integration
Detection alerts delivered directly into any WhatsApp group — with snapshot image, camera name, class, and confidence score
Live Feature

Phylax Vision can post detection alerts directly to any WhatsApp group — reaching every member instantly with no additional app or account required on their side. Each alert includes the AI-annotated camera frame with the detected object's bounding box drawn in, so the recipient sees exactly what was detected and where.

→ AI-annotated snapshot — the detection frame with bounding box is sent as an image in the chat
→ Camera · Class · Confidence — camera name, detected object class, and AI confidence percentage in every message
→ Multiple groups per site — link a site to more than one WhatsApp group; each group configurable independently
→ Image on/off per group — control whether the snapshot is included or text-only for bandwidth-sensitive recipients
→ Configured in the mobile app — link and manage WhatsApp groups directly from phylax.vision.app; no server access needed
→ No recipient app needed — anyone in the WhatsApp group receives the alert; no phylax.vision account required
Ideal for
Neighbourhood watch groups Armed response operators Family & household sharing Estate security teams Property managers
South Africa context: WhatsApp is the dominant communication channel for security response in South Africa — neighbourhood watches, armed response companies, and estate management all coordinate primarily on WhatsApp. Phylax Vision meets them where they already are.
👥
Phylax.vision
3 members
Phylax.vision WhatsApp group — real detection alerts with AI-annotated snapshots
Actual detection alerts — live Phylax.vision WhatsApp group
Detection → Response pipeline
AI detects person
→
Workflow triggers
→
📱 Push notification
💬 WhatsApp group
💡 Lights on
🚨 Siren
🏢 Control room API
All responses configurable per camera, per detection class, per time window — in the visual workflow editor.
Roadmap

Detection is live.
Deterrence and recognition are next.

Phylax Vision already sees and alerts. The next releases make it act — speaking to intruders before they move, and recognising every vehicle at the gate. All delivered over-the-air to every deployed device.

🗣️
In Development
AI Voice Deterrence
When the AI detects a person, the phylax device speaks a warning through the camera's own speaker or a wired on-site speaker — before anyone touches your property. Messages are written by you and context-aware: "You in the blue shirt at the front gate — you are being recorded."
English Afrikaans isiZulu More by config
No cloud round-trip at detection time — instant response
🚗
In Development
License Plate Recognition — ANPR
A dedicated LPR node reads number plates at gates and driveways — day and night. Whitelisted plates can open the gate automatically; blacklisted or unknown plates trigger sirens, WhatsApp alerts, or your control room. Full searchable plate history, managed from the same dashboard and app.
Whitelist / Blacklist Gate automation Plate history
Separate hardware SKU — same platform, provisioning, and OTA updates
⚡
Planned
Deeper Automation
An expanded visual workflow editor with typed hardware outputs — relays, sirens, gates — so a detection, a spoken warning, and a physical response chain together in one workflow.
Visual editor Relay outputs
Builds on the workflow engine already shipping today
Same device. Same platform. Every capability above ships over-the-air to devices already in the field — that is the point of owning the full stack.
Market Opportunity

A massive market with a clear gap.

South Africa is one of the world's largest per-capita consumers of private security services. AI-powered surveillance is replacing legacy CCTV — and the incumbents are cloud-dependent, expensive, or require camera replacement.

R57B[5]
SA private security industry (annual)
>9,000[6]
registered security companies in SA
$50B+[7]
global AI surveillance market by 2030
The Problem
Existing solutions require cloud
Cloud-dependent AI means ongoing bandwidth costs, privacy risks, POPIA exposure, and single points of failure. Rural and load-shedding environments are completely underserved.
"Rip and replace" kills adoption
Most AI surveillance products require new proprietary cameras, alienating the millions of properties with existing camera infrastructure already in place.
No professional-grade local platform
Security installers and alarm companies in SA have no locally-built, locally-supported AI security platform to resell. They depend on imported hardware with no software ownership.
Our Advantage
Works with existing camera infrastructure
No camera replacement required. The phylax device integrates with IP cameras already installed — the most important feature for installer-led sales in the upgrade market.
Full stack ownership
We own every layer from firmware to mobile app. No SDK dependencies, no per-device licensing to a third party, no platform lock-in. 100% margin on software.
B2B reseller channel ready
Designed to be sold through alarm companies and security installers who already have billing relationships, installation teams, and thousands of existing CCTV clients.
Why Invest

Built to own the
AI security stack.

4
Fully shipped products
4
AI model tiers (edge + cluster)
0
Cloud dependency for AI
∞
Devices scalable per server
💰
Revenue model — three streams
① Hardware device — once-off licence fee ② Annual AI model subscription per device ③ Enterprise API & monitoring station access
01
Recurring Revenue Model
The OTA update infrastructure creates a natural subscription model: devices receive AI model upgrades, binary patches, and new features centrally managed through the server. SaaS revenue without cloud infrastructure costs — the AI runs on the customer's hardware.
02
Scalable Without Marginal Cost
Each new device deployment adds zero incremental cloud AI processing cost. The server manages configuration, updates, and connectivity — all heavy lifting happens on the edge. Gross margins improve as the device base grows.
03
B2B Channel — Security Companies
South Africa has over 9,000 registered security companies with existing CCTV client bases hungry for AI upgrade paths. Alarm companies and security installers are ready-made distribution partners — they own the billing relationship, we supply the AI platform.
04
AI That Evolves With the Threat
Because phylax.vision trains its own AI models in-house, detection accuracy improves continuously. As criminal tactics evolve — new vehicle types, concealment methods, timing patterns — our models can be retrained and deployed to all devices via OTA. Competitors relying on static third-party AI cannot match this adaptability.
05
Proprietary AI — Device-Locked
Our AI models are custom-trained and cryptographically protected. Each deployed model is locked to the specific device it was licensed for — it cannot be copied, redistributed, or run on unauthorised hardware. This creates a defensible moat that is extremely difficult for competitors to replicate.
06
Clear Enterprise Expansion Path
The existing multi-site architecture, cloud integration workflows, and role-based access system position phylax.vision to move up-market into enterprise campuses, municipalities, and commercial property management groups — without re-engineering the core platform.
07
Privacy-First Regulatory Position
POPIA compliance is built into the architecture — no footage leaves the customer's premises. As AI surveillance regulation tightens globally, phylax.vision's edge-first design becomes a competitive requirement, not just a differentiator.
Investors & Partners

The platform is built.
The market is ready.
Let's talk.

We are raising to accelerate distribution through South Africa's R57B private security industry and to fund further AI model development. Book a 20-minute call or request the full investor deck below.

Contact details available on request.
References & Sources
  1. "98% of Security Camera Alarms Are False — Here's What That's Actually Costing You." Intellisee, 2024. intellisee.com. Source for the 98% false alarm rate, the $1.8 billion annual US cost, and the 36 million false alarm calls figure.
  2. "Armed Response Prices in South Africa — 2026 Guide." Armed-Response.co.za, 2026. armed-response.co.za. Source for the R300–R550 per false call-out charge range; R400+ used here as mid-range estimate.
  3. "AI Motion Sensors: The End of the False Alarm Era." Bay Alarm Company, 2026. bayalarm.com; and "AI Motion Detection vs. PIR: What's the Difference?" Zumimall, 2024. zumimall.com. Sources for the sub-5% false positive rate of AI object detection vs. 80–98% for legacy motion sensors.
  4. "Edge vs. Cloud Camera Processing: Latency, Privacy and Cost." AI USB Cam, 2024. aiusbcam.com. Source for the 500–700 ms cloud AI latency figure vs. under 50 ms for on-device edge processing.
  5. South African Private Security Industry — Annual Market Size. Security Association of South Africa (SASA) and industry reports, 2024–2025. The R57 billion figure is widely cited across SA business and security industry publications as the annual revenue of the private security sector. Readers are encouraged to verify current figures at sasa.org.za.
  6. PSIRA Annual Report — Registered Security Businesses. Private Security Industry Regulatory Authority (PSIRA), South Africa. psira.co.za. PSIRA is the statutory regulator of the private security industry in South Africa and publishes registered entity counts annually. The 9,000+ figure reflects active registered security service providers.
  7. "AI Surveillance Market Size, Share & Trends Analysis Report to 2030." Grand View Research / MarketsandMarkets, 2024. Multiple independent market research firms project the global AI-powered surveillance market to exceed $50 billion by 2030, driven by edge computing adoption and the replacement of legacy CCTV systems. Figures vary by methodology and scope; $50B+ represents a conservative mid-range projection across available reports.
  8. Standard NVR motion detection vs. AI object classification. CCTV Camera World, "Smart Motion Detection Explained," 2024. cctvcameraworld.com; and Sirix Monitoring, "AI vs. Motion Detector — What's the Difference?" 2024. sirixmonitoring.com. Both sources confirm that basic/legacy NVR systems use pixel-change motion detection, not object classification. Note: mid-range NVRs (including some Hikvision and Dahua models) now include on-board AI — the table refers to the basic/legacy category.
  9. Milestone XProtect and Genetec Security Center — analytics architecture. Milestone Systems, "Understanding Intelligent Video Analytics," official documentation. milestonesys.com; and Genetec, "On server or edge — how to choose the right video analytics," official blog. genetec.com. Both platforms support edge and server-side AI analytics. The table marks this as partial (~) because full AI capability requires enterprise licensing not typically accessible to SME or residential markets.
  10. Hikvision AcuSense NVR — third-party camera compatibility. Hikvision official product page, DeepinMind Edge series. hikvision.com. AcuSense NVRs run AI on limited channels with third-party cameras; full-channel AI capability requires Hikvision AcuSense cameras. DeepinMind Pro explicitly supports non-AI cameras across all channels. The table reflects AcuSense as the representative product line.
  11. Ring and Arlo — paid subscription required for AI detection features. Ring official support: "Ring Protect Subscription Plans." ring.com/support/plans. Arlo official knowledge base: "Animal, Vehicle, and Person Detection — Arlo Secure." kb.arlo.com. Both Ring and Arlo explicitly require a paid subscription plan for AI person/vehicle/object detection. Basic motion detection works without a subscription; AI classification does not.
Phylax Vision own system data: The detection count (93,228), uptime (47 hours), and camera count (5 cameras) displayed in the Live Deployment section are captured directly from a running phylax.vision deployment as of June 2026. These are real operational figures, not projections or estimates.