A camera can verify some lockout-tagout steps — that a lock and tag were applied, that a panel was not opened before them — but it cannot verify that the energy is actually isolated, which is the step a meter and a human confirm. So the honest role for vision in LOTO is evidence, not enforcement: it produces a timestamped record that the procedure ran in the right order, while the try-step that proves zero energy stays human.
That boundary is the whole design. LOTO is a sequence, and the dangerous failure is doing it out of order or skipping the verification. A camera is good at witnessing sequence and presence. It is blind to voltage. Build the system around what it can witness and it is a strong control; pretend it confirms isolation and it is a dangerous one.
The LOTO steps, split by who can verify them
| Step | Camera-verifiable | Record-only | Human-only |
|---|---|---|---|
| Energy sources identified | — | — | Yes — judgment and knowledge |
| Isolation device operated | Partially — switch position in view | — | Confirm correct device |
| Lock applied | Yes — lock present on device | — | — |
| Tag applied | Yes — tag present and legible-at-scale | Logged | — |
| Verify zero energy (try-step) | No | — | Yes — meter and attempted start |
| Second-person check | Partially — second person present | Logged | Confirm the check happened |
| Panel opened | Yes — opened only after lock/tag in record | — | — |
Read down the camera column: it covers the before-and-after of the dangerous act and the presence of the controls. The one step it cannot touch is the one that actually keeps a person alive — verifying zero energy. That stays with a meter and a qualified person, every time.
Evidence, not enforcement — and why that is the right frame
The tempting pitch is "the camera stops the panel from opening". That is the wrong frame and it fails on contact with a works council and with reality: a vision layer on existing CCTV does not physically interlock a panel, and selling it as enforcement oversells it and understates the human's role.
The right frame is evidence. The layer produces a record that, for this intervention, a lock and tag were in place before the panel opened and a second person was present. After an incident or during an audit, that record answers "was the procedure followed" with a timestamp instead of a recollection. It supports the discipline; it does not replace the permit-to-work and the qualified person who own the isolation decision.
Regulatory framing, named only
Singapore's Workplace Safety and Health (General Provisions) Regulations cover the duty to control hazardous energy during maintenance, and Malaysia's Factories and Machinery Act 1967 with DOSH guidance covers the equivalent isolation duties. This page names those instruments and invents no clause numbers. Both regimes expect a documented, followed isolation procedure; a per-intervention event record showing lock, tag and sequence is evidence that the procedure ran, sitting under the permit system, not over it.
Where this fits, honestly
Hypernology's AI Safety layer runs a context-aware vision-language model on existing CCTV, deployed in about an hour, which is why it reads sequence and presence rather than requiring new interlocking hardware. The step-by-step verifiability table is reasoning about what is observable from a frame; the deployed proof is context-aware safety monitoring on existing cameras generally, not a published LOTO case. Treat the table as the specification to test on your own panel-room footage, not a claimed result.
Related reading: the CDM 2024 Malaysia AI safety monitoring post, the machine-guarding compliance post, the hot-work fire-watch monitoring post, and /solutions/hyperq-ai-safety.
Frequently asked questions
Can a camera confirm lockout-tagout isolation? No. It can verify that a lock and tag were applied and that the panel opened only after them, but it cannot confirm zero energy. The try-step with a meter stays human.
Is this enforcement or evidence? Evidence. A vision layer on existing CCTV does not physically interlock a panel; it records that the procedure ran in order, which supports audits and incident review but does not replace the permit-to-work.
Which steps still need a person? Identifying energy sources, confirming the correct isolation device, the zero-energy verification, and owning the second-person check. The camera records presence and sequence around these; the judgment is human.
Send one panel-room's camera view and your LOTO procedure, and within two weeks we will return which steps your sightlines can evidence and which stay human-verified, before any rollout. Start at https://apac.hypernology.net/contact.