The original Scarecrow documents preserve valuable ideas, but they also contain historical prices, ambitious targets and legal assumptions. This page shows what we can responsibly say now and what a real pilot must still establish.
AfriForum describes this as its own civil-society research and notes possible underreporting and differing definitions. It is not an official SAPS national ground-truth dataset. Read the 2025 report.
This is an official historical census figure, not a current addressable-market count and not proof that every unit needs or can afford Scarecrow.
What these figures establish: rural violence and security risk justify investigation. They do not establish that robots will prevent an attack, reduce response time, save money or suit every farm.
The historical 70-90% aspiration has no field denominator or controlled comparison and is not a public product claim.
Rural conditions vary. A pilot must measure each site's existing and tested alert-to-human workflow.
Old hardware prices, exchange rates and monthly bands are planning history—not quotations.
Evidence may support future underwriting discussions; only an insurer can decide cover, terms or premium effects.
Mapping PSiRA, POPIA, PAIA, safety and SACAA questions is not approval or certification.
Force-capable concepts in early documents are excluded from v1 and from the current software scope.
One word cannot honestly describe both build progress and proof quality. Public records use a lifecycle status, may add a dependency flag, and state the evidence qualifier separately.
The lab gateway already exports a structured incident evidence ZIP with a per-file SHA-256 manifest, and that path passes an automated test. Human-readable reporting, media evidence, signatures, trusted timestamps, retention and controlled release workflows remain planned.
| Evidence family | Records | Decision supported |
|---|---|---|
| Availability | Gateway uptime, device heartbeat, connectivity and power interruptions | Whether the configuration can be relied on during agreed hours |
| Detection | Controlled scenarios, alerts, classifications, false alerts and missed observations | Whether a device adds useful information in that environment |
| Human response | Review time, acknowledgement, escalation and closure | Whether the workflow helps rather than overloads people |
| Safety | Stops, no-go tests, degraded states, faults and near misses | Whether physical operation may progress |
| Privacy | Camera zones, access, exports, retention and requests | Whether data collection remains proportionate and governable |
| Maintenance | Interventions, downtime, parts, updates and recovery effort | Whether the system is supportable and what it truly costs |
| Economics | Actual hardware, installation, connectivity, support and incident inputs | Whether a lease, shared cluster or partner model is viable |
Links establish the context stated beside them. They do not endorse Project Scarecrow or substitute for deployment-specific advice.
