ISO 45001 and Video Evidence: What Camera Records Can—and Cannot—Show
Where camera event records support ISO 45001 §6.1.2, §7.5, §9.1.1 and §10.2 — and what a signature, an event count or a quiet month cannot establish.
- workplace-safety
- iso-45001
- documented-information
- audit
A camera can record a workplace safety event. Whether that record supports an ISO 45001 audit depends on what was detected, how the record was preserved, and what your organisation did next.
Most industrial sites already have cameras. Some use them for security, some for operational visibility, others to investigate incidents after they happen.
But having footage is not the same as having reliable evidence that supports a safety management system.
ISO 45001 does not require CCTV or AI video analytics. It requires a systematic approach to occupational health and safety: hazard identification, risk assessment, operational controls, performance evaluation and continual improvement.
Camera-generated event records can support that approach. They cannot replace it.
Does an ISO 45001 auditor need video footage?
Not necessarily.
An auditor evaluates whether your occupational health and safety management system meets the applicable requirements and operates effectively. That can mean reviewing documented information, interviewing workers, observing activities and examining how incidents and corrective actions are handled.
Video may be relevant, but ISO 45001 does not prescribe it as a mandatory evidence format. Footage can itself be documented information when it is properly managed and controlled. The practical difficulty is that hours of video are hard to search, interpret, verify and connect to specific safety decisions.
A structured event record is different. Instead of asking someone to review an entire recording, it identifies a specific detection: the rule that produced it, the camera and zone, the recorded time, and any supporting image or evidence.
The key question is not simply whether an event was captured.
It is whether the organisation can explain the event, assess its reliability and show what happened next.
Already using Axis cameras? Check what your cameras can support — six questions, under a minute, no sign-up.
Where camera-generated records fit ISO 45001
Four clauses are especially relevant.
| ISO 45001 clause | What the organisation must address | How camera records may help |
|---|---|---|
| 6.1.2 — Hazard identification and assessment of risks | Identify hazards and assess OH&S risks | Observations from predefined monitored zones |
| 7.5 — Documented information | Create, maintain and control appropriate information | Identifiable records, access control, retention and change history |
| 9.1.1 — Monitoring, measurement, analysis and performance evaluation | Monitor OH&S performance with appropriate methods | Event counts, trends, and what is known about the limits of the measurement |
| 10.2 — Incident, nonconformity and corrective action | React, investigate, act and review effectiveness | Support for the investigation, and a before-and-after comparison of the corrective action |
These are supporting uses, not automatic evidence of compliance.
1. Hazard identification: §6.1.2
Consider a warehouse where pedestrians occasionally walk into an area used by forklifts.
The organisation has identified that interaction as a hazard, assessed the risk and put controls in place. A camera-based system configured to recognise people entering a defined vehicle zone adds observations: when and where that defined situation is actually detected.
A camera does not identify every workplace hazard, judge how severe an injury could be, or carry out the risk assessment. A detected event is an input to professional judgement, not a substitute for it.
2. Documented information: §7.5
An exported clip or event log can support the documented-information requirements when it is suitably identified, protected and controlled. Useful attributes of a record include:
- a unique identifier, the camera, the zone and the recorded time
- the detection rule and the relevant configuration version
- where the record came from, and any available integrity verification
- access restrictions and an audit history
- a retention and disposal policy
- the organisation's own review, approval and follow-up, where applicable
Integrity matters most. Prova, for example, creates cryptographically signed event records on compatible Axis cameras. Verification shows whether a signed record has been changed since it was created.
But a valid signature has a specific meaning.
It establishes the record's integrity and where it was signed. It does not establish that the detector classified the situation correctly, that the camera clock was accurate, or that every hazardous event was captured.
Those questions need their own evidence. A report can be tamper-evident without being a complete or infallible account of site conditions.
3. Monitoring and measurement: §9.1.1
Traditional safety reporting mostly counts injuries and incidents that have already caused harm. Event-based monitoring adds another view: how often predefined, potentially hazardous situations are detected before anyone is hurt. Defined, validated and interpreted properly, these observations can be useful leading indicators.
But the denominator matters.
Say a monitored zone produces 12 detections in one 100-hour observation period and 8 in the next. That is a reduction in recorded detections. It is not, on its own, a sign that workplace risk fell by a third. The reading depends on whether:
- the camera was working and the zone stayed in view
- detection rules and sensitivity stayed comparable
- false positives and missed detections were taken into account
- working hours, pedestrian traffic and vehicle activity were comparable
- gaps in coverage and changes in configuration can be explained
A period with zero events is not a zero-risk workplace.
ISO 45001 also asks that monitoring and measuring equipment be calibrated or verified as appropriate, and used and maintained properly. Where a camera system produces OH&S measurements, the organisation should set suitable checks for that method. For video analytics that can mean camera availability, field of view, clock synchronisation, rule versions and periodic checks of detection performance.
A system health report helps. It does not, by itself, validate the accuracy of the analytics.
4. Incident investigation and corrective action: §10.2
Suppose a pedestrian walks into a vehicle lane and the system records the event. The record helps an investigator establish the time, place and conditions of that detection.
A responsible person then reviews it. Was the detection correct? Why did the situation occur? Were the existing controls inadequate? Is corrective action needed?
The organisation might change the traffic layout, add a physical barrier, revise a procedure or retrain a shift. After the change, comparable observations help assess whether the intervention had the intended effect.
Corrective-action effectiveness still has to be judged in context. A lower event count alone does not show that every relevant risk is adequately controlled. The decision, the action and its effectiveness remain part of the organisation's OH&S management system.
Five questions an auditor — or an HSE manager — should ask
Before relying on automated camera events as supporting evidence, ask:
1. What exactly does the system detect? Are the detection rules explicit? Do they describe the relevant conditions? Can a responsible person tell a confirmed hazardous situation from an automated detection?
2. Can the record's integrity be checked? What integrity mechanism is there, and what data does it protect? How is the signing key tied to the camera, and what is needed to verify the record later?
3. How reliable was the observation? Was the camera working? Was its view blocked? Were configuration changes recorded? Are clock accuracy, missed detections and other limits known?
4. Is the record managed properly? Who can access it? How long is it kept? Are changes logged? Can the organisation retrieve the evidence when it is needed?
5. What did the organisation do about it? Was the event reviewed? Was it relevant to an identified hazard or incident? Was action needed, and was its effectiveness assessed?
An event log supports these questions. It does not answer all of them for you.
Workplace privacy is part of the design
Workplace video analytics also has to be assessed against privacy and employment law.
Under the GDPR, where personal data is processed, the organisation needs a lawful basis, a defined purpose, transparency, proportionality, appropriate safeguards and retention controls. A data protection impact assessment (DPIA) is required when the processing is likely to result in a high risk to people's rights and freedoms; whether that threshold is met depends on the actual processing and the applicable guidance.
Anonymisation, or the absence of names, does not automatically make information anonymous. People can be identified indirectly, for example by combining event times and locations with other records.
National rules can add requirements. In Slovenia, Article 78 of ZVOP-2 restricts video surveillance inside workplaces and requires the employer to consult the representative trade union and the works council before introducing it.
Worker participation and trust belong before deployment, not only at audit time. Technology meant to support workplace safety must not quietly become a tool for scoring individual productivity or for unrelated monitoring of employees.
How Prova approaches safety evidence
Prova is built for compatible Axis cameras that already run object analytics. Rather than adding a second detection model, it uses the camera's own analytics and applies your event rules on the device. It keeps three functions apart:
Detection. The Axis camera provides people and vehicle analytics for the zones and rules you configure.
Evidence. Prova creates a signed event record on the camera, with the event's metadata. Depending on your policy, one still image anonymised on the camera accompanies the event, or the record carries no image at all.
Review. The evidence registry verifies every signature and produces audit packs for your own review and documentation. Your operators mark each incident reviewed, confirmed or dismissed, with a note; the review sits beside the signed record and never changes it.
Prova never uploads, stores or streams footage. It does not identify individual workers, and it does not decide corrective actions. It is not a safety interlock and does not replace workplace safety controls.
The audit pack also lists every camera at the site: for how much of the period it was reporting, its longest gap, the configuration version it was running, and its licence and image health. Where something was not reported, the pack says so, rather than letting a gap pass for normal conditions.
For sites whose cameras are not Axis, VAP takes a different approach: it works with any camera that has an RTSP stream and a second, lower-resolution stream, runs detection on a box on site or on your own server, and keeps metadata and a short evidence clip for every alarm, plus the operator's decision once it is reviewed. Its evidence model is not the same as Prova's camera-signed records.
What a useful ISO 45001 audit pack should show
A practical safety evidence report keeps five layers apart:
- Observed detections — what the configured system recorded.
- Evidence integrity — whether the records can be verified.
- Observation limits — what is known about camera availability, configuration and detection reliability.
- Human review — which records were examined and what was concluded.
- Management action — what the organisation decided, carried out and later evaluated.
The layers matter because cryptographic integrity and management-system effectiveness are not the same thing. A trustworthy report also states what it cannot establish.
The camera records. The organisation remains responsible.
ISO 45001 is not about collecting as much footage as possible. It is about identifying hazards, managing risks, evaluating results and improving occupational health and safety.
Camera-generated event records can make part of that process more consistent, traceable and easier to review. They cannot replace the people, procedures and decisions an effective management system depends on.
The strongest safety evidence is not simply a record that cannot be altered. It is a record whose origin, limits, interpretation and resulting action can all be explained.
See what your cameras can support
Without talking to anyone: run the camera compatibility check, and read the sample Prova audit pack (PDF), generated from synthetic data, to see its structure, verification details and stated limits.
With our team: we confirm your cameras and firmware, set up your zones and rules, and agree the goals before anything starts. See Prova — the Starter package covers one site, up to four cameras, for 60 days.
References
- ISO 45001:2018 — Occupational health and safety management systems
- ISO 45001:2018/Amd 1:2024 — Climate action changes
- EDPB Guidelines 3/2019 on processing of personal data through video devices
- EDPB — data protection guidance for SMEs, including DPIAs
- Slovenian Personal Data Protection Act (ZVOP-2)
- Information Commissioner of Slovenia — guidelines on video surveillance under ZVOP-2