Sightproof · for security and safety teams · works with the cameras you have

    Every camera alert comes with proof.

    What was detected, a second look with its reason, and your team's verdict — with faces blurred before anything is stored.

    Alert Center: a warehouse camera frame with numbered detection boxes and a forklift-lane zone, the second look verdict with its reason and what was visible, and the Confirm / False positive buttons
    Sightproof screen · illustrative data · AI-generated scene

    Cameras you can trust.

    Most camera AI raises more alerts than anyone can read. Sightproof checks each one, shows its work, andlearns from the people who know your site.

    Second look

    Every alert checked before anyone is woken up.

    A vision model looks at each alert frame and says whether the box really shows what the detector reported — with its reason and what it saw. A gate post is not a person, and now the queue says so.

    • •Verified, likely false or unclear — never a silent dismissal
    • •Shadow mode: the second look never changes or mutes an alert by itself
    • •Muting stays locked until your own verdicts prove it right
    Second look on a night loading-dock frame: three detections verified with reasons, one marked likely false because the box shows a gate post, and the agreement card with muting locked
    Sightproof screen · illustrative data · AI-generated scene
    Privacy at the source

    Faces are blurred before a frame is ever stored.

    Sightproof blurs faces the moment a frame is captured, so alert images, the second look, search and training only ever see blurred faces. If blurring cannot run, the frame is not kept.

    • •Unblurred originals encrypted, kept only for alert frames, deleted on your schedule
    • •Viewing an original needs an admin and a written reason — and lands in the audit trail
    • •No face recognition, no face matching, ever
    Privacy settings: faces blurred before storage, encrypted originals kept seven days, a reason-required view of the original and the audit trail entry it creates
    Sightproof screen · illustrative data · AI-generated scene
    Workplace Safety pack

    Turn the same cameras on your safety rules.

    Tell the cameras what matters in plain words — hard hats, high-vis vests, people in forklift lanes — with no training to start. Findings get the same second look and your team’s verdict.

    • •Safety score and shift compliance computed from verified findings
    • •“No data” when nothing was measured — never a perfect score by default
    • •Early access: we scope it to your sites and measure it on your footage
    Workplace Safety: plain-word labels such as hard hat and high-vis vest, a warehouse frame with checked PPE, a safety score from verified findings and shift compliance showing no data where nothing was measured
    Sightproof screen · illustrative data · AI-generated scene
    Incidents that close

    From alert to a documented decision.

    Group verified alerts into an incident, give it an owner, and keep every note and verdict in one activity log. An incident cannot be closed without saying how it was resolved.

    • •Owners, site admins, operators and viewers — access by site
    • •Every action recorded: verdicts, assignments, notes, original-frame views
    • •Escalation when an alert is not acknowledged
    Incident workflow: an owner, the verified alerts in the incident, an activity log with notes, and a resolve action that requires a resolution note
    Sightproof screen · illustrative data · AI-generated scene

    One engine for any camera question

    Nothing is dismissed without a person, and nothing is claimed until your own verdicts measure it.

    1 · Detect

    Any label, no training to start

    Point the detector at what matters in plain words, or use a model tuned to your site.

    2 · Second look

    Checked, with a reason

    A vision model confirms or rejects each alert and says what it saw.

    3 · Your verdict

    Your team is the answer key

    Confirm or reject; agreement with the second look is measured, not assumed.

    4 · Your model

    Better at your scenes

    Verified frames from your own cameras fine-tune a detection model for your site.

    Packs

    Security pack

    Available
    • Restricted-area and after-hours alerts by camera zone
    • Incidents with an owner, an activity log and a required resolution note
    • Escalation when an alert is not acknowledged
    • Shared queues by site

    Workplace Safety pack

    Early access
    • Hard hats, high-vis vests and no-go zones from plain-word labels
    • Safety score and shift compliance from measured findings
    • Same second look and human verdict as security alerts
    • Warehouses, plants and construction sites

    What we won't do

    Camera AI lives or dies on trust. Here is where Sightproof draws the line.

    Identify people

    No face recognition and no face matching. Faces are blurred before a frame is stored.

    Dismiss an alert on its own

    The second look only labels. A person confirms or rejects, and muting stays locked until your verdicts prove it right.

    Claim accuracy we have not measured

    No vendor benchmark numbers. Agreement is measured against your own team’s verdicts, on your cameras.

    Keep a frame it could not protect

    If face blurring is required and cannot run, the frame is not stored — and the camera shows why.

    Built on the RiseLab platform

    The same platform RiseLab uses to fine-tune and run AI on a customer's own data.

    Your model, your data

    Fine-tune detection on frames from your own cameras, using the confirmed and false-positive verdicts your team records on alerts.

    Cloud or on your premises

    Detection runs in RiseLab's cloud. Cameras on a private network connect through an edge device on your premises. The second look is off until an admin turns it on.

    Audit & access

    An audit log of alert, incident and privacy actions, permission-based access by site, and scheduled and on-demand reports.

    Book 15 minutes about your cameras

    Tell us about your sites and the alerts your team ignores today. We will show the second look, privacy and the agreement card, and scope a pilot measured against your own verdicts.

    Prefer email? Tell us about your sites and we will propose a time.

    See it on your own cameras

    Bring one site. In a pilot you see every alert with its second look, faces blurred before storage, and how often the second look agrees with your team — measured, not promised.