The camera continuously holds the last period of video and commits it when recording starts - so the footage includes what happened before the button was pressed.

Incidents escalate before anyone decides to record, and the beginning is usually the most contested part.

Buffer length is configurable, typically from a few seconds to a minute or more.

Whether the buffer includes audio is a separate setting and often a legal question, because recording audio before any notification is given is treated strictly in some jurisdictions.

A longer buffer consumes more battery, so it is a trade rather than a free feature.

Set it in the policy, not on the device, so that it is consistent across every camera and defensible if challenged.

Through a docking station that charges and uploads simultaneously into a managed evidence store with an audit trail.

Manual transfer by cable or memory card is unworkable at any scale and breaks the chain of custody.

Docking stations upload automatically, so the process depends on the camera being returned rather than on anyone remembering to copy files.

The evidence store must record who accessed, viewed, exported or redacted each recording, and when.

Originals must not be editable. Redaction for disclosure produces a new copy with the original preserved.

Categorising footage at the point of upload - incident, routine, evidential - is what makes retention manageable later.

As short as the purpose allows, with an automatic deletion policy and a mechanism to hold specific footage for longer.

Most routine footage has no ongoing purpose after a short period, and keeping it creates both cost and risk.

Data protection regimes generally require a stated purpose and a defined retention period rather than indefinite storage.

Footage relating to an incident, a complaint or legal proceedings has to be held beyond the routine period, which requires an explicit hold mechanism.

Deletion should be automatic. A manual process is not applied consistently and cannot be evidenced.

Set the periods in policy with legal input and configure the system to enforce them rather than relying on procedure.

They differ by jurisdiction, and audio recording is treated far more strictly than video in many territories.

Most regimes expect people to be told they are being recorded, either by visible signage, a marking on the camera, a verbal announcement, or all three.

Some require explicit consent for audio, and a few prohibit covert audio recording outright regardless of the video position.

Recording in private spaces - homes, changing areas, medical settings - is restricted or prohibited in many places.

Employment law may also apply where the camera records colleagues as well as members of the public.

Take local legal advice per territory and write the answer into the operating policy rather than leaving wearers to judge.

When recording starts and stops, what is announced, what must not be recorded, and what happens to the footage - and it has to be trained, not just published.

Continuous recording for a whole shift creates enormous volumes and captures a great deal that has no purpose.

Event-based recording, started by the wearer at defined trigger points, is the usual approach and needs those trigger points defined clearly.

The policy should state what the wearer announces, and when, so notification is consistent.

It should also state what must not be recorded and what to do if recording begins somewhere it should not.

Audit compliance. A policy nobody follows does active harm, because it creates an expectation the footage will exist when it does not.