Onboarding video
Heart Rate from Video
Pulse recovered from facial colour — proof a live person is present, not a photograph or a replayed recording.
TrustShield OSINT
AML Intelligence
Money laundering moves an estimated 2–5% of global GDP a year, and the controls meant to catch it still ask whether a document is valid rather than whether the person holding it is real, present and acting freely. FaceOff answers the second question, at onboarding and on every session after it.
Where it plugs in
- Onboarding — E-KYC and video KYC
- Transaction monitoring, continuously
- High-risk sessions, re-checked passively
- Cross-institution models, without pooling data
OutputIdentity assurance a deepfake cannot pass, and a smaller alert pile
- Of global GDP laundered every year — USD 800 billion to 2 trillion
2–5%
Of global GDP laundered every year — USD 800 billion to 2 trillion
UN estimate
- CAGR on the global AML market, taking it from $4.13B in 2025 to $9.38B by 2030
17.8%
CAGR on the global AML market, taking it from $4.13B in 2025 to $9.38B by 2030
- Regimes tightening at once: the EU's AMLA in 2025, the U.S. AML Act of 2020, and India's PMLA extended to digital assets
3
Regimes tightening at once: the EU's AMLA in 2025, the U.S. AML Act of 2020, and India's PMLA extended to digital assets
Where it plugs in
Three layers of an AML programme
FaceOff does not replace an AML stack. It answers the one question the stack cannot: whether the human on the other end is real, present and acting of their own accord — at onboarding, during every session after it, and across institutions that are not allowed to compare notes.
Customer due diligence
Document KYC is slow and forgeable, and plain facial recognition now loses to a deepfake. Liveness, synthetic-ID detection and stress analysis close the gap at the point the identity is first asserted.
- Liveness
- Deepfake & synthetic ID
- Emotional stress
Transaction monitoring
Rule-based monitoring generates alerts that go nowhere and misses coordinated activity spread thin across many accounts. Behavioural signals catch a mule account being operated by someone other than its holder.
- Behavioural biometrics
- Continuous authentication
- Periodic liveness checks
Collaborative intelligence
The schemes span institutions; data protection rules stop those institutions sharing customers. Federated learning trains one shared model across all of them without any raw data moving.
- Federated learning
- Secure aggregation
- Privacy-enhancing technologies
The base layer is what makes the two above it better than they could be alone. A mule pattern seen at one institution improves the model every participant runs, and none of them ever sees another's customers.
What it verifies
Four signals behind a KYC decision
Each of these runs on the onboarding video or the live session, and each has its own page on the platform.
The video feed itself
Deepfake Detection
Artefacts of AI-generated or manipulated media, so a synthetic applicant cannot pass a video KYC.
Micro-expression
Facial Emotion Recognition
Duress — the signature of someone being coached or coerced into opening a mule account.
Voice
Audio Tone Sentiment
Vocal stress that contradicts what is being said, on the same call it is being said on.
The market
Five forecasts, one direction
Independent projections of the AML market, none of which agree on the number. The spread is the useful part: every one of them compounds at double digits into the early 2030s.
- Core AML market by 2030, up from $4.13B in 2025 — 17.8% CAGR
$9.38B
Core AML market by 2030, up from $4.13B in 2025 — 17.8% CAGR
- An alternate forecast for 2030, from $1.73B in 2024 — 16.2% CAGR
$4.24B
An alternate forecast for 2030, from $1.73B in 2024 — 16.2% CAGR
- Software and services by 2032, from $4.48B in 2024 — 14.8% CAGR
$13.56B
Software and services by 2032, from $4.48B in 2024 — 14.8% CAGR
- The optimistic scenario for 2032, from $3.29B in 2023 — 19.2% CAGR
$19.05B
The optimistic scenario for 2032, from $3.29B in 2023 — 19.2% CAGR
- The software-only segment by 2032, from $2.04B in 2023 — 12.6% CAGR
$5.91B
The software-only segment by 2032, from $2.04B in 2023 — 12.6% CAGR
Growth on this scale is driven by three things at once: regulation tightening, digital finance expanding, and criminal tactics moving faster than the rule-based systems built to catch them. Figures as published in FaceOff's AML briefing; the laundering estimate above is the UN's.
Why it is getting harder
AML has stopped being a line item
It is a strategic exposure now — institutions that lag risk penalties, reputational loss and systemic vulnerability. The forces driving the spend are the same ones making the job harder.
- Regulatory pressure
- FATF sets the global standard, the EU's AMLA launches in 2025, the U.S. AML Act of 2020 tightened reporting and corporate transparency, and India's PMLA now reaches digital assets and fintech platforms.
- Regulatory fragmentation
- Those regimes do not agree with one another. A cross-border institution runs several definitions of the same obligation simultaneously.
- Privacy against intelligence
- The schemes span institutions; data protection rules stop those institutions comparing notes. That contradiction is the one federated learning exists to resolve.
- Tactics that compound
- Synthetic identities, deepfaked video KYC, coerced mule accounts and DeFi rails — each of them newer than the rule-based system meant to catch it.
- False positives are the real cost
- Rule-based monitoring buries analysts in alerts that lead nowhere, while the coordinated activity spread thin across many accounts goes unflagged.
- Digital expansion
- Online transactions, digital banking, crypto and cross-border trade have moved AML out of the back office and onto the board agenda across banking, BFSI and government.
The rest of the line
More in TrustShield OSINT
Where it runs
Industries deploying AML Intelligence
Each sector page covers the threat model, the controls, and the regulators that apply.
AML has shifted from compliance to strategy.
Identity assurance a deepfake cannot pass, behavioural monitoring that shrinks the alert pile, and a shared model no one has to hand their customers over to build. See it against your own onboarding flow.