A gate your agent calls before it acts, and a console your team works from.
Decisions, reviews, incidents, policies and the evidence behind each one, in one place.
The console shows what every agent was asked to do, what Verbrisk decided and why, across every device and channel you run.
Product illustration with sample data.
01 · Trust gate
One call before every action. Verbrisk runs the four checks, weighs them against your policy for that action, and returns a decision your agent can act on in milliseconds.
hostile < 0.33 · suspect < 0.66policy v12 · critical × hostile → block
02 · Review queue
Every flagged command becomes a work item. Triage it, investigate it, resolve it as an attack, a false positive or a scheduled test, and keep the record.
03 · Policies
Decide what the agent does for every combination of action and trust level. Dry-run a change against last week's traffic before you publish it.
| Action | Trusted | Unverified | Suspect | Hostile |
|---|
04 · Fleet analytics
Carriers that repeat across devices, sources that cannot be checked, a spike at one location. Fleet analytics turns single decisions into a picture of what is happening.
05 · Live monitor
A live spectrogram of any enrolled device, scored one second at a time. Useful for testing a room, a device or a suspected emitter.
roomspeech only · nothing above 18 kHz
06 · Audit trail
Every decision is written with its inputs, scores, thresholds, policy version and outcome. Export it, stream it to your SIEM, or attach it to an incident.
Voice is personal data. Verbrisk is built to keep it in your control, and to keep test traffic out of your records.
Hosted in your chosen region. Audio is scored in memory; retention is off unless you turn it on.
Deployed in your VPC. Models and policies run beside your agents; nothing leaves your account.
The signal check is small enough for a hub or gateway, where the full-rate audio still exists.
Next step
A pilot measures false-accept and false-reject rates on your devices and channels, with your policy.