newCallback verification is a control that voice cloning quietly broke.
Proveniro
Banks & payments

The controls held for thirty years. Generative AI broke one of them last year.

AI voice cloning did not create a new attack surface so much as invalidate an old assurance. Proveniro restores it — inside callbacks, contact centres, onboarding and the AI agent traffic now arriving on all three — without asking you to redesign the processes your auditors already signed off.

ai verdict latency
< 400 ms
call path change
none
time to shadow
days
pricing
per verification
Exposure

Three places an AI-cloned voice converts directly into loss.

Each of these is a control designed around a human ear, and each is now defeated by an open-weight AI model that costs nothing and takes minutes to use.

Payment authorisation callbacks

Vendor bank changes and high-value transfers are confirmed by phone. An AI agent answers in the voice of someone the approver trusts, holds a fluent conversation, and every downstream control shows green.

what it costs
Single-incident losses in the millions, with a recovery window measured in hours.

Contact centre authentication

Knowledge-based questions are already compromised by breach data. Voice was the last human backstop, and your agents cannot hear the difference between a customer and a model — nobody can.

what it costs
Account takeover at scale, plus the reputational cost of the disclosure.

Onboarding & KYC

Injection attacks feed an AI-generated face straight into the capture pipeline, never passing a real camera. A synthetic customer becomes a permanent mule account.

what it costs
Regulatory exposure, remediation programmes, and money-laundering liability.
Capabilities

One AI verification layer across every channel that carries a voice or a face.

Deployed passively, surfaced in the tools your teams already use, and reversible at any point.

Callback verification

Live scoring of the callback leg with a verdict returned inside the conversation, wired to the payment hold.

Contact centre scoring

Passive AI analysis of inbound calls with a risk badge and model reason codes in the agent desktop or CRM.

Onboarding liveness

Injection detection, presentation attack detection and generative artefact scoring on KYC captures.

AI agent verification

Allow your own enrolled voice agents through on a signed, scoped credential while unregistered synthetic callers are held or blocked.

Regulator-ready evidence

Every AI verdict sealed with inputs, model versions and thresholds, exportable for examination and AI Act reporting.

Deployment controls

Regional residency, configurable retention, SSO and SCIM, and a contractual no-AI-training guarantee.

Rollout

From first call to enforced decision in about a month.

Nothing is automated until you have seen the numbers on your own traffic.

  1. 01

    Scope

    Identify the payment bands, call types and onboarding flows where exposure is concentrated.

    week 0

  2. 02

    Fork media

    SIPREC or WebRTC mirror from the existing platform. No change to the call path or IVR.

    week 0–1

  3. 03

    Shadow score

    Every eligible call scored, no action taken. Base rates, precision and recall established.

    weeks 1–2

  4. 04

    Enforce

    Hold on high-confidence synthetic verdicts, review queue for the ambiguous middle.

    week 3–4

Assurance

Built to survive an examination, not just a demo.

Financial services buyers do not need to be sold on the AI threat. They need to know the control is defensible, measurable, governable and reversible.

the honest version

No AI detection system is perfect, and any vendor telling you otherwise is selling you a number rather than a control. What we will commit to is this: measured model performance on your own traffic before you automate anything, an operating point you choose, and a record of every decision that you can defend afterwards.

what you get
integrations
Genesys, Amazon Connect, NICE, Five9, Twilio, Avaya, generic SIPREC and webhook
decision surface
Agent desktop badge, case management, payment hold API, webhook to your orchestrator
evidence
Sealed per-decision record with inputs, AI model versions, thresholds, timings and actor trail
residency
EU, UK and US regions, selected per tenant, with in-region processing and storage
retention
Zero-retention scoring available; otherwise configurable up to 90 days
assurance
SOC 2 Type II in progress, penetration test summaries, EU AI Act documentation, DPIA and DPA templates
commercials
Per-verification pricing with committed-volume tiers; enterprise licence available
Get started

Find out what your callbacks are currently missing.

A two-week shadow evaluation on your own traffic, scoped to the payment bands where your exposure is concentrated.

soc 2 type ii in progress · eu & uk data residency · no ai training on your data