KPMG Buys Into Deepfake Detection as Testing Labs Multiply

Consolidation says the fraud is permanent

Strategic investors do not buy into categories they expect to shrink. The consultancy's own research context makes the thesis explicit: liveness detection software alone is insufficient against sophisticated bypass attacks, so detection has become a standing capability that enterprises must operate, staff, and audit. That is precisely the shape of spending consultancies monetize.

The demand data supports it. Deepfake documents account for 80.10 percent of AI-enabled fraud attacks in this year's ring data, deepfake identity fraud was projected to grow nearly five-fold in 2026, and the detection market has passed the multi-billion mark with more than six billion checks performed industry-wide. Every one of those numbers is an argument for detection as a line item rather than a pilot.

Accredited testing changes the buyer's question

Until recently, a vendor's deepfake defense was as good as its demo. Injection attack detection accreditation changes that: standardized laboratory testing of whether a system catches virtual cameras, stream substitution, and synthetic media injection, performed by certified independent labs. KPMG's own Swiss lab recently gained exactly this accreditation, and the lab field is expanding as providers line up to be tested.

For buyers, the practical upgrade is a sharper question set. Not "do you detect deepfakes," but: which accredited lab tested your injection defense, on which OS versions, and what did it miss? Certification is a floor, not a ceiling, presentation attack standards famously lagged the injection era, but a floor with third-party evidence beats a ceiling made of slideware.

The deeper evaluation still goes beyond any lab visit: continuous adversarial pressure against the live stack. deepidv runs Arbiter against its own deployments with current generator tooling and fraud persona kits precisely because certification snapshots age, while deepeye's structural liveness is designed to hold on physics rather than on the artifacts labs tested last quarter.

What the maturation means for the market

Three consequences follow. Detection pricing firms up: infrastructure with auditors attached commands enterprise contracts, not pilot budgets. Standalone detectors face a squeeze: as consultancies and platforms internalize detection, point products either integrate into verification stacks or become features. And claims deflate to evidence: between accredited labs and adversarial testing, the gap between marketed and measured performance is becoming visible to buyers for the first time.

The winners in that market are the stacks that were built to be tested, where detection is one layer of a verification architecture with liveness, capture forensics, and provenance correlation underneath it, and where the red-team results are a sales asset rather than a secret.