Catch the callers your fraud checks are letting through
Flag cloned voices and AI voice agents about twelve seconds into the call. Send the flag to the systems your team already works in. Nothing changes on your call path.
Synthetic.
The attacks
The call sounds like your customer.
It isn't.
A cloned voice asks your care team to move a number to a new SIM.
1,055%
rise in unauthorised SIM swaps reported in the UK in 2024, to nearly 3,000 cases.
Source: Cifas, Fraudscape 2025 · May 2025 ↗Example call
“I've lost my phone. Can you move my number to a new SIM today?”
A caller with your customer's voice passes security and takes over the account.
$40bn
in US fraud losses that generative AI could enable by 2027, up from $12.3bn in 2023 (forecast).
Source: Deloitte Center for Financial Services · May 2024 ↗Example call
“I'm locked out. Can you reset my password and update my phone number?”
A policyholder's cloned voice calls to send a claim payout somewhere new.
£638,000
of claims fraud found every day in 2025 by one UK insurer, with more claims backed by AI-generated images and documents.
Source: Aviva · June 2026 ↗Example call
“About my claim: can the payout go to a different bank account?”
AI voice agents call in at scale, looking for the agent who will say yes.
62%
of organisations surveyed had a deepfake attack using social engineering or automated processes.
Source: Gartner survey of 302 security leaders · Sept 2025 ↗Example call
“Hi, I'd like to check the balance on the account ending 4471.”
Measured live
A verdict at twelve seconds.
Then it keeps listening.
19 / 22
Fakes caught live
AI voice agents and cloned voices, over internet and phone-network calls
0 / 12
Genuine calls flagged
Real callers on clean lines
0 ms
Added to your calls
Trueline listens to a copy, so no call waits
How it works
It listens to a copy of the call.
Your calls never wait for it.
01Connect your phone line
Send Trueline a copy of each call from Twilio, Telnyx or your own phone system. Your call routing and agent tools stay exactly as they are.
02We listen as the caller speaks
Trueline checks the voice every two seconds and makes its first decision twelve seconds into the call. It keeps listening until the call ends.
03The flag reaches your systems
The moment a synthetic voice is detected, your systems are told, with the reason. Your team decides what to do; Trueline never ends a call.
Trueline only listens to a copy of the call, so it can never delay or drop one.
Audio is never stored
It lives in memory for up to eight seconds while it is scored, and is never written to disk. No voiceprint is made, so there is nothing to enrol and nothing to leak.
It cannot touch your calls
Trueline reads a copy, the way a call recorder does, so nothing is added to the call. If we stop mid-call, your call carries on and the record says how much was scored.
Where it runs
Run it where your calls are.
The same build either way: nothing in Trueline depends on a particular cloud, so your own box is the hosted service with your keys.
The questions we get on the first call.
Every answer here is what the system does today. Where the honest answer is a weakness, it is written as one.
See how it works →01What about calls from mobile phones?
Our numbers come from internet and carrier routes, which is where voice agent fraud arrives first and where Trueline is strongest.
A mobile leg is a harder path for any detector. The call is compressed by the handset and the mobile carrier before anyone else hears it, and that first hop takes away part of what makes a synthetic voice detectable.
It is where our measurement work is now, and we will not tell you it is solved. If your callers are mostly on mobiles, say so before a pilot rather than after, and the report will tell you where your own traffic stands.
How we measure it →02If a call is flagged, do you stop it?
No. Trueline is never in the call path, so it has nothing to stop. The flag leaves as a signed webhook and your policy decides what happens next: step up the ID check, hold the payment, move the call to a specialist.
It is an assist for the people who already make that call, not an automatic block.
03Do you need a voiceprint of our customers?
No. Nobody is enrolled and no biometric template is created or kept. Trueline does not compare the caller to a known voice. It judges how the speech itself was produced.
So it works on the first call from a number nobody has heard before, and there is no voice database to govern.
04What if the attacker uses a voice tool you have never seen?
There is no list of tools to keep up to date, so a new engine is not a new signature to add. What gets scored is the way the speech was made, not which product made it.
That is not the same as saying nothing gets through. Where a route does defeat it, as a mobile handset in the path does today, we publish that rather than patch around it.
05Can we see why a call was flagged?
Yes, and it is the same record every time: which rule fired, how many four-second windows crossed the threshold, how much of the caller's speech had actually been heard, and the model version that decided.
The reason travels with the flag and stays on the call record, so the decision can still be explained months later.
Point one week
of calls at it.
You get back every flag with its reason, every fake it missed, and the routes it was measured on. Your calls, not our bench.
Or see how it works first →