Competitor Analysis¶
Face recognition for security operations — the landscape around Patrol Sight · HumanityAI'd — July 2026
Confidential
Download
1. The one-line finding¶
The face-recognition market is crowded at the fixed camera (NEC, IDEMIA, Corsight, the Chinese majors) and at the cloud API (Clearview and its imitators). It is close to empty at the patrol-worn edge: the only vendors shipping glasses-based recognition are Chinese (LLVision, Rokid) and effectively unsellable to internationally exposed GCC projects, while the one regional precedent — a Dubai integrator's 50-unit smart-glasses deployment in 2019 — proved the concept and then went quiet. Nobody currently combines AR-glasses capture + fully on-premises GPU inference + a customer-owned watchlist + GCC-compliant data residency. That combination is our position.
2. Landscape at a glance¶
| Vendor | Format | Deployment | Wearable | GCC footprint | NIST-ranked |
|---|---|---|---|---|---|
| HumanityAI'd Patrol Sight | AR glasses | On-prem | Yes — core product | Target | Not yet — planned |
| NEC (NeoFace) | Fixed cameras / gates | On-prem, enterprise | Mobile app, no glasses | Strong (DXB smart gates with emaratech) | Yes — perennial top-2 in FRTE 1:N |
| IDEMIA | Gates / borders / ID | On-prem, gov-grade | No | Strong (UAE eBorders; Zayed International; KSA MoUs with TCC and SAMI, 2025) | Yes |
| Thales | Borders / ID | On-prem | No | Long-standing Gulf presence (FR-specific contracts to confirm) | Yes |
| Corsight AI (Israel) | Fixed + edge | On-prem + edge | No native glasses | Marketing to ME agencies; 50+ countries | Claims top Western rank |
| Oosto (ex-AnyVision, Israel) | Fixed + edge | On-prem/edge | No | Historic Gulf activity (to confirm) | Yes |
| SAFR (RealNetworks, US) | Camera-embedded | Cloud or fully offline edge | No | To confirm | Yes |
| Herta Security (Spain) | Fixed / crowd | On-prem/edge | No | Regional exhibitor (to confirm) | Yes |
| Clearview AI (US) | Cloud API over 30B+ scraped images | Cloud only | No | None reported | — |
| Hikvision / Dahua / SenseTime / Yitu (China) | Cameras + analytics | On-prem | No glasses | Large installed CCTV base | Restricted |
| LLVision (China) | Police glasses (GLXSS Pro) | Edge + tablet | Yes | China only | — |
| Rokid (China) | AR glasses + FR packages | Edge | Yes | China police use reported | — |
| Vuzix / RealWear (US) | Wearable hardware platforms | Needs third-party software | Yes (hardware only) | Vuzix Blade used in the UAE deployment below | n/a |
| NNTC — iFalcon Face Control Mobile (Dubai) | Smart glasses + offline database | Edge, offline | Yes | UAE — 50 units in security operations (2019) | To confirm |
3. The competitors that matter¶
NEC and IDEMIA — the incumbents who own the buyer relationship¶
Both are deeply embedded in GCC biometric infrastructure: NEC behind Dubai airport smart gates with emaratech, IDEMIA behind the UAE's multi-biometric entry/exit programme, Zayed International's biometric journey (over a million passengers processed), and 2025 memoranda with Saudi partners TCC and SAMI Advanced Electronics. They are also the NIST accuracy leaders — NEC has ranked top-two in all eight FRTE 1:N categories.
Where they beat us: algorithm pedigree with independent NIST validation, government-scale references, decades of procurement muscle, and existing contracts we would be bidding beside.
Where we beat them: neither ships a patrol-worn product. Their architecture assumes the subject walks to a fixed point — a gate, a kiosk, a camera. Patrol Sight assumes the officer moves and the subject does not cooperate. Our answer to their strength is not to claim better matching; it is to occupy a different position (mobile, hands-free, sub-100 ms in-lens) and to plan our own NIST FRTE submission so the accuracy conversation stops being asymmetric.
The Chinese majors and LLVision/Rokid — the only shipping wearable precedent, and why it does not close the seam¶
Chinese police smart glasses are the real proof this works: at Zhengzhou East Railway Station in February 2018, railway police wearing LLVision glasses matched faces against a 10,000-suspect database in roughly 100 milliseconds and caught 7 fugitives and 26 identity-fraud travellers during the Lunar New Year rush. Rokid glasses have been used on Beijing and Shenzhen patrols.
Why the seam stays open: Hikvision, Dahua, SenseTime and Yitu — together roughly a third of the global video-surveillance market — sit on the US Entity List over Xinjiang human-rights abuses. For airports, international events, sovereign-fund projects and anything with Western financing or partnership exposure, that is a procurement and reputational problem regardless of price. LLVision and Rokid have no GCC presence, no regional support, and no data-protection story built for PDPL.
Our answer: identical capability class, neutral supply chain, on-premises architecture, and a governance posture we publish rather than avoid.
NNTC / iFalcon — the closest comparable, and our best evidence¶
A Dubai-based integrator deployed 50 Vuzix Blade smart glasses in UAE security operations in 2019, scanning crowds against an offline database of violators, missing persons and suspects, with a 2020 release scaling to hundreds of wearables alongside thousands of fixed cameras over 4G/5G.
Why this is good news, not bad: it proves a GCC security buyer has already bought exactly this concept, in this region, with an offline database — the objection "nobody does this here" is already answered. It went quiet, and the hardware generation it was built on (Vuzix Blade, 2019) is far behind current AR glasses.
Where we beat it: current-generation glasses with hardware H.264 at 1440p, a measured sub-100 ms pipeline, a 94,000-person GPU index with a documented path to 13 million, a two-layer tracker that makes continuous streaming affordable, and a full command console with audit and tracking history rather than a wearable checking a list.
Corsight and Oosto — the Western challengers with a regional complication¶
Both are technically credible on-premises and edge recognition vendors, and Corsight markets actively to Middle East homeland-security agencies. Oosto's story carries additional weight: after raising USD 352 million it was acquired for USD 125 million in January 2025 at roughly six times revenue — a signal that standalone face-recognition software struggles to sustain a premium without a hardware or workflow anchor. Both are Israeli-origin, which is a procurement complication in parts of the GCC that varies by country and should be confirmed case by case rather than assumed.
Where we beat them: neither has a native wearable product, and both sell software into someone else's camera estate. We sell the capability an officer actually wears, with the hardware, the pipeline and the console as one deployable system.
Clearview AI — the cautionary tale we position against¶
A cloud API over 30 billion-plus scraped web images, with a USD 51.75 million US class-action settlement approved in March 2025 and contested regulatory enforcement in the UK. It is unusable for GCC data-residency buyers and it is the single best contrast for our governance pitch: the customer's own watchlist, enrolled with a recorded lawful basis, held on the customer's own server, searchable only by named operators, every query logged. Where Clearview's model is "we already have everyone", ours is "you have exactly who you are lawfully entitled to have".
4. The commodity floor and the pricing lesson¶
Standalone face-matching software has weak pricing power — Oosto's exit multiple demonstrates it, and cloud APIs price recognition by the call. Our pricing power comes from the combination nobody else assembles: the wearable form factor, the measured latency, the on-premises sovereignty, and the operational envelope (deployment tooling, boot recovery, audit, console). Sell the deployed capability, never the matching algorithm.
5. Five differentiators to lead every conversation with¶
- Recognition where the officer is — not at a gate, not in the control room, not after the encounter. The only format that identifies someone while the officer is still standing in front of them.
- Measured, not claimed — ~68 ms typical glasses-to-label, published stage by stage, with a 94,000-person index and a documented path to 13 million. Nobody else in this category publishes a latency budget.
- Sovereign by construction — the watchlist, the faceprints, the alerts and the audit trail never leave the customer's network. Cloud vendors cannot follow, and GCC data-protection law increasingly requires it.
- The customer owns the watchlist, and a human owns the decision — no scraped database, a recorded lawful basis per entry, mandatory human verification before any action, and a complete audit trail. This is what makes the deployment defensible to a regulator.
- Neutral supply chain — not Chinese-sanctioned, not cloud-dependent, and without regional sourcing sensitivities. For an internationally exposed project, that alone shortlists us.
6. Objections this landscape will raise (and our answers)¶
| Likely objection | Answer |
|---|---|
| "NEC/IDEMIA already supply our biometrics." | They supply gates and cameras — fixed points the subject walks to. We supply recognition that walks with your officers. We deploy beside them, not instead of them. |
| "What's your NIST ranking?" | We don't have one yet, and we won't imply otherwise. What we do have is a fully measured end-to-end pipeline in a real deployment, and a plan to submit to NIST FRTE. Ask the NIST-ranked vendors for their glasses-to-label latency budget. |
| "The Chinese vendors do this for less." | They do — and they're on the US Entity List, have no GCC presence or support, and no PDPL-aligned data story. On an internationally exposed project that's a procurement problem, not a price comparison. |
| "Someone already tried smart glasses here in 2019." | Correct, and it worked — which is our best evidence. That was a 2019 device with an offline list. We bring current glasses, a measured sub-100 ms pipeline, a 94k-person GPU index and a full command console. |
| "Isn't facial recognition legally risky?" | The risk lives in scraped databases, cloud processing and automated action. We have none of those: your watchlist, your server, your operators, human verification mandatory, everything logged. We'll support your DPIA. |
| "Can you match at 50 metres across a plaza?" | No — reliable to about 5 metres, 6–7 in good light, 8–10 in long-range mode. Anyone promising a plaza from a head-worn camera is selling you a fixed-camera problem. |
| "You're a small vendor — can you support a national deployment?" | Not alone, and we don't claim to. We bid with a licensed regional integrator for installation, fleet operations and first-line support; our engineering plus their operational muscle. |