Safe
Safe is a Spanish computer-vision platform that connects to existing IP and CCTV cameras to detect workplace hazards in real time — missing PPE, falls, vehicles, restricted zones, smoke — sending instant alerts and generating automatic audit reports.
What is Safe?
Safe is an occupational health and safety platform built on computer vision, published by SIALI TECHNOLOGIES SL in Santander, Cantabria, Spain. Rather than requiring new hardware, it connects to the IP and CCTV cameras a site already owns: it detects cameras automatically, speaks RTSP and GenICam, and is claimed compatible with more than 95% of the IP cameras produced worldwide.
The platform advertises more than 25 detectable risk types. Nine use cases are included in the base licence: complete PPE, presence or intrusion in a forbidden zone, vehicles outside their authorised area, forklift-to-person distance, excessive forklift speed, distance from suspended loads, a person falling, fire or smoke, and prolonged immobility. Additional vision-language (VLM) modules cover narrower cases — an unfastened vest, speciality PPE such as hair nets, beard covers and welding masks, unauthorised opening of doors or furnaces, a raised tipper, obstacles on evacuation routes. Bespoke projects go further still: models built from scratch, deep integration with dashboards, ERP (SAP) and SCADA, leak detection in ATEX zones, big-bag alignment, overhead-crane safety through PLC coordinates, and ergonomics and posture analysis.
A second product line, Safe Care, targets hospitals and residential care homes: falls in beds and corridors, disoriented residents, staff PPE, smoke, restricted zones, distance between people and obstacles.
Privacy is built into the pipeline. Faces are blurred irreversibly the moment the image is acquired, digital silhouettes are available, and captured video stays on the customer's own servers. Deployment is either fully on-premise or hybrid, with edge servers per site and a central point in the cloud. The stack is containerised microservices — Docker Compose on Linux behind Traefik, or native Windows — with PostgreSQL, local S3-style object storage, MediaMTX for HLS and WebRTC, a REST API, WebSockets, a message queue and a rules orchestrator. Detection runs on a YOLO engine with pose analysis and tracking.
The publisher claims more than 200,000 people protected and more than 50,000 risk events detected, and states a 0.4-second detection cycle in production projects; these are vendor figures. Published testimonials come from Grupo Red Electrica (electrical substations) and Acciona (construction). The site is available in Spanish and English, and Safe is delivered as a web application backed by on-site inference servers. Logos shown on the site for Microsoft, Canon, NVIDIA, Asepeyo, Prevcontrol, Lisdata and Ironsys are partners and clients, not certifications or product integrations.
What it does
- Continuously analyse the live feeds of a site's existing IP cameras over RTSP and GenICam
- Detect missing or badly worn PPE: helmet, hi-vis vest, gloves, glasses, harness, mask, hearing protection
- Detect falls, people on the ground and prolonged immobility
- Detect restricted-zone intrusion, vehicles outside their authorised area, forklift-to-pedestrian distance, excessive forklift speed and distance from suspended loads
- Detect smoke, fire and water leaks
- Trigger responses automatically: audible alarms, light panels, SMS, email, and commands sent to PLCs or forklifts
- Produce automatic daily, weekly and monthly reports and audits, and compute a Safe Score whose target is zero
When to use Safe / When not to
A quick filter to help you decide if Safe is the right fit.
When to use Safe
- Occupational health and safety (PRL/HSE) managers in factories and industrial plants who need continuous, evidence-backed monitoring of risk situations
- Site supervisors and construction project managers on works where PPE compliance and vehicle movements are hard to police manually — Acciona is a published reference
- Warehouse and logistics managers watching forklift-to-pedestrian distance, forklift speed and obstructed evacuation routes
- Security managers of electrical sites and substations, where restricted-zone intrusion is the core risk — Grupo Red Electrica is a published reference
- Directors of hospitals and elderly care homes, through the Safe Care line: falls in beds and corridors, disoriented residents and staff PPE
When not to use Safe
- Organisations with no cameras on site and no hardware budget: quotes never include cameras, servers or PCs, which must be sourced through integration partners
- Very small companies or single sites on a tight budget: the recurring floor runs into several thousand euros a year, on top of a EUR 1,500 proof of concept
- Anyone who needs to identify individuals by name: faces are blurred irreversibly at capture by default
- Military, citizen-surveillance or rights-suppression use cases, which the acceptable use policy forbids outright
- Teams with no IT resource, or anyone looking for a mobile app, a browser extension or a self-service public API: none of these exist, and GPU servers, RTSP streams, ports 80/443/22, NTP and camera placement all have to be managed
How to use Safe
A typical end-to-end flow, from setup to results.
- Book a 30-minute meeting through the demo page, or try the online demo on your own cameras
- Ask for the free feasibility analysis before committing to any bespoke development
- Optionally start with a proof of concept: 1 to 5 cameras on one area, one basic use case, EUR 1,500 as a one-off payment
- Check the camera prerequisites: 720p minimum (1440p recommended), RTSP, H.264/H.265, at least 10 FPS (15 recommended), 4-6 Mbps per camera
- Install the on-site inference server with an NVIDIA GPU, or choose the hybrid model; the Safe team performs the installation over SSH or VPN, or on site
- Connect the cameras: Safe detects the existing CCTV cameras automatically and attaches to their streams
- Define the zones to monitor, then select which situations to detect and how severe each one is
- Configure the rules, thresholds and responses: audible and light alarms, SMS, email, PLC commands
- Follow the real-time dashboard (HLS/WebRTC), the full event history and the Safe Score
- Generate the daily, weekly and monthly reports and track the resolution tasks
Pros & Cons
Pros
- Reuses the cameras already installed: no fleet replacement, with claimed compatibility above 95% of IP cameras
- A public price grid and an online estimate calculator, which is rare in this market
- Detailed public technical documentation: architecture, security, AI accuracy, business continuity and hardware prerequisites
- A published subprocessor table naming each provider and its function
- Irreversible face blurring from the moment of acquisition, video kept on the customer's own servers, and a fully on-premise option with no cloud dependency
- Nine use cases included in the base licence with no activation surcharge, and a EUR 1,500 one-off proof of concept to validate before committing to licences
- Named industrial references (Grupo Red Electrica, Acciona), plus a corporate guarantee under which a bespoke development stays active for the whole group, only the licence being repaid
Cons
- No hardware included: cameras, servers and PCs remain the customer's responsibility, sourced through integration partners
- High entry budget — EUR 2,500 to EUR 5,000 a year for the core licence, EUR 30 to EUR 35 per camera per month at the first tier, EUR 500 to EUR 1,500 of setup — and every published amount is an indicative MSRP range with no contractual value before technical validation
- No public API documentation, no mobile application and no browser extension
- No customer DPA is published, and Safe claims no certification of its own: the ISO 27001, SOC 2, HIPAA and PCI DSS references belong to its subprocessors
- No retention period is published for the product's event data; the privacy policy only covers the website
- Accuracy depends heavily on installation quality: 3 m mounting height, 0-45 degree tilt, minimum pixel densities, penalising occlusions; VLM modules are announced at only above 85% accuracy, extreme accuracy being a bespoke project billed by engineering day
- Site and interface limited to Spanish and English, technical documentation only in Spanish, and inconsistent support details: the documentation points to addresses on a third-party domain
Pricing & Plans
There is no permanent free plan. A free trial is offered in the form of an online demo run on the customer's own cameras, alongside a free feasibility analysis. The cheapest paid entry point advertised is a proof of concept at EUR 1,500 as a single one-off payment, covering 1 to 5 cameras, one area and one basic use case, with no initial recurring licence. Beyond that, the mandatory recurring floor is the SAFE-Start core licence at EUR 2,500 to EUR 5,000 per year, to which per-camera licences are added from EUR 30 to EUR 35 per camera per month at the first tier, plus EUR 500 to EUR 1,500 of setup and integration. A one-camera subscription therefore costs approximately EUR 2,860 in licences for the first year, before setup. Hardware is never included, and all amounts are published as indicative ranges.
- POC — EUR 1
- 500 one-off payment
- 1 to 5 cameras
- one area
- one basic use case
- no initial recurring licence
- Licencia base core "SAFE-Start" — EUR 2
- 500 to EUR 5
- 000 per year
- mandatory
- includes the 9 basic use cases
- Camera tier Basico — 1 to 25 cameras
- EUR 30 to EUR 35 per camera per month
- Camera tier Estandar — 26 to 75 cameras
- EUR 24 to EUR 28 per camera per month
- Camera tier Premium — 76 to 150 cameras
- EUR 19 to EUR 22 per camera per month
- Camera tier Corporativo — 151 cameras and above
- EUR 18 to EUR 20 per camera per month
- Ad-hoc VLM modules — EUR 900 to EUR 1
- 800 of setup per case
- plus EUR 200 to EUR 400 per month per case
- Bespoke projects and developments — priced by engineering day
- Onboarding & setup — EUR 500 to EUR 1
- 500 with existing IP cameras
- or EUR 990 to EUR 1
- 990 for a project without cameras or with mobile units
Data, GDPR & hosting
A consolidated view of how Safe handles your data.
GDPR overview
Implementation is unusually concrete. A dedicated document, "Marco de Seguridad, Privacidad y Proteccion de Datos (GDPR)", describes irreversible face blurring applied the moment the image is acquired: processing and storage use the blurred image and the original face cannot be recovered. Full-body blurring is available. Controls include AES-256 at rest, TLS 1.2+ in transit, multi-tenancy with per-organisation file partitioning, and RBAC (User, Admin, Super Admin). Events can be exported and erased for data-subject rights. The privacy policy cites Regulation (EU) 2016/679, keeps a record of processing activities, routes requests to info@sialitech.com, allows complaints to the Spanish AEPD, and frames non-EEA transfers through adequacy decisions, appropriate safeguards or BCRs. Subprocessors are listed. Missing: no customer DPA, and Safe claims no certification of its own; the SOC 2, ISO 27001, HIPAA and PCI DSS references belong to its subprocessors.
Who owns the data?
Safe publishes no customer licence agreement and no product terms of sale, so data ownership is addressed only in the site FAQ and the technical documentation. The FAQ states that images captured in production are stored locally on the customer's own disks or servers, and that the customer manages that data. The GDPR document adds strict multi-tenancy: every camera, alert, user and recording is tied to a unique Tenant ID, files are physically partitioned per organisation, and events can be exported or erased. The legal notice covers only the website's own content, owned by SIALI TECHNOLOGIES SL. The acceptable use policy's clause on data ownership is an internal policy addressed to the publisher's staff and contractors, not a customer clause.
Reuse rights
The customer hosts and manages the event images itself, and the system allows events to be exported and deleted in support of access and erasure rights. Daily, weekly and monthly reports can be generated and customised by the customer, and the full history is presented by Safe as irrefutable evidence in the event of a dispute. No published clause explicitly authorises or forbids the customer from reusing or republishing that data outside the system: with no product terms of sale on the site, reuse is left to whatever the customer negotiates. In a fully on-premise deployment, third-party subprocessors are replaced by the customer's own internal services.
Data retention & training
Hosting summary
The publisher is established in Spain, so the platform sits under Spanish and EU law. Three deployment modes are documented: fully on-premise, with a dedicated server per site so that AI processing and video storage stay inside the customer's private network; hybrid, with edge servers per plant plus a central point in a private or public cloud; and cloud. The FAQ confirms local or cloud deployment depending on the customer's requirements, and the site repeats throughout that captured video stays on the customer's servers. Storage relies on PostgreSQL for configuration and history and on local S3-style object storage for event images and evidence. For the SaaS and demo plan, the disclosed subprocessors are Cloudflare (CDN/WAF), Resend/SMTP (alert emails), PostHog and Vercel Analytics (web demo only) and Sentry (error monitoring); in a fully on-premise deployment these are replaced by the customer's internal services. The privacy policy allows transfers outside the EEA for some providers, framed by adequacy decisions, appropriate safeguards, BCRs or Spanish authority approval. Published RPO/RTO figures range from 5 minutes to 1 hour, with an annual recovery test. No hosting country or region is named anywhere on the site.
Where Safe works
Country-level availability.
Not available in
Things to keep in mind
Risks and trade-offs to weigh before adopting Safe.
- Continuous video monitoring of employees can drift towards productivity control or individual sanction even with faces blurred; the publisher's own blog acknowledges that six in ten Spanish companies have not handled employee resistance to AI, which makes a collective agreement before deployment essential
- A reversible blurring mode exists and can, once notified and enabled beforehand, let an administrator reveal faces: a capability that needs to be tightly framed
- Detection quality depends entirely on installation: a badly placed camera produces false negatives, and therefore a false sense of safety
- The Safe Score can become an end in itself and push teams to under-report incidents
- The round-the-clock history is presented as proof in a dispute; it can be used in both directions, against the employer as well as against the employee
- With no customer DPA and no security certification of Safe's own, the GDPR impact assessment stays entirely on the customer's side
- Delegating human vigilance to the machine is a real temptation, since the tool only covers what a camera can see, and the entry cost plus a possible CAPEX commitment of one to five years adds lock-in
Setup & Integrations
Technical difficulty
Daily use is easy: the FAQ states that no prior knowledge is required, and defining zones, situations and severities is presented as a matter of minutes. Deployment is another matter. It needs a dedicated server with an NVIDIA GPU sized to the camera count (from an RTX 5060 for a proof of concept up to eight NVIDIA L4 for 50-120 cameras), 32-256 GB of RAM, 4-120 TB of video storage, a PoE switch, 4-6 Mbps per camera, ports 80/443 and 22, and NTP. Safe's team installs over SSH or VPN, or on site; no hardware is supplied.
Deployment
Integrations
Supported languages
Behind Safe
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
How do I know whether Safe is compatible with my existing installation?
Is the system hosted on the internet?
How fast does Safe detect and analyse?
Are the captured images and videos protected?
Which risks can Safe detect?
How much does it cost?
Is there a free trial or a free plan?
Is there a mobile application?
What hardware do I need?
Are there any usage restrictions?
Should you pick Safe?
Safe is a vertical tool, tightly specialised in industrial occupational risk prevention and in the care of vulnerable people. Its main strength is that it works with the camera fleet a site already owns, removing the usual capital barrier to computer-vision safety. Around that, SIALI TECHNOLOGIES SL publishes material that is unusual in this market: full technical documentation covering architecture, security, AI accuracy and business continuity, a subprocessor table, and a public price grid with an online calculator. Irreversible face blurring at acquisition and a fully on-premise option make it credible for works councils and data protection officers.
The weaknesses are just as clear. The entry ticket is high — a mandatory core licence of EUR 2,500 to EUR 5,000 per year before a single camera is licensed — and hardware is excluded from every quote. There is no public API documentation, no mobile application, no certification of Safe's own (the SOC 2 and ISO 27001 badges cited belong to its subprocessors) and no published retention period for the product's event data.
The publisher is a modest Spanish company based in Santander with real industrial references, Grupo Red Electrica and Acciona, and vendor-claimed figures of more than 200,000 people protected and 50,000 risk events detected. Value depends heavily on camera placement and server sizing: a badly placed camera produces false negatives and a false sense of safety. The acceptable use policy also carries an explicit ethical and geographic restriction, notably on Israel. Safe suits organisations that already own cameras, have IT resources and a real safety budget, not small sites looking for a cheap plug-and-play answer.
- Choosing a selection results in a full page refresh.
- Opens in a new window.