Deepfake detection evolving beyond onboarding into continuous financial trust

As forecasted in Biometric Update’s 2026 Deepfake Detection Market Report, deepfake detection continues to evolve from an onboarding tool to a system integrated across the lifecycle. “Static, one-time authentication to access enterprise or consumer accounts has become obsolete,” the report says. “Banking, fintech, and high-value gaming platforms are hemorrhaging money to postlogin fraud. They need a system that continuously evaluates risk throughout the entire user session.”

New products and resources from IngenID and Resemble AI, alongside new guidance from JPMorgan and Accenture, reinforce the significant market shift from point-in-time identity verification toward continuous protection. In effect, as deepfakes and the techniques to use them for fraud expose more vulnerabilities across the customer journey, deepfake detection has gained the dimension of time: not just checking that the person who arrives at the point of access control is real, but making sure they stay that way while they’re inside.

IngenID levels up to continuous voice authentication 

An update to IngenID’s Twilio connector adds continuous, real-time deepfake detection to its existing voice biometric authentication and classification system. A release from the voice security firm says the upgraded connector lets contact centers verify callers and provides assurance they’re dealing with real callers throughout the entire conversation – not just at login.

Deploying natively on Twilio, IngenID’s connector streams audio continuously via Twilio Media Streams to re-verify caller identity throughout an interaction. It also flags synthetic or manipulated audio as it happens, and makes continuous protection viable at high call volumes without the pricing overhead associated with legacy systems.

“Fraud doesn’t happen at the same moment of every call,” says Nick Bartolotti, CTO at IngenID. “It can happen at any time. Attackers know a one-time check only protects the start of a call, so we made authentication continuous: our Twilio connector now listens for a live, consistent voice for the full duration of the call, catching a synthetic voice swap, a hijacked session, or a fraudulent handoff the instant it happens.”

ResembleAI resource unpacks deepfakes for AML compliance

A new resource on Resemble AI’s blog explores how deepfake detection is being integrated into AML compliance workflows.

“Deepfake detection for AML compliance is becoming an important part of modern risk management,” it says. As onboarding, authentication and interaction go remote, synthetic media can be used for identity fraud at a greater number of pressure points – driving the need for continuous monitoring.

Deepfake detection, says the blog, “may contribute to ongoing monitoring by supporting high-risk account verification, account recovery processes, authentication events and sensitive transaction approvals.” It thus becomes part of an architecture that complements existing AML frameworks, and also includes biometric verification, risk-based authentication and other controls.

That said, Resemble AI also recognizes that these relatively recent market mutations are currently loose in an unstable regulatory landscape. “Expectations around synthetic media risks and deepfake detection continue to evolve. Financial institutions may need to adapt their controls as regulatory guidance becomes more defined.”

‘A static defense posture is a losing one’

An article co-authored by executives from J.P. Morgan Payments and Accenture argues that effective defense against deepfake fraud now requires “a layered technology approach that meets the threat at the critical window between when a payment is initiated and when it’s executed.”

“This window is shrinking: as payment rails get faster, the opportunity to recall funds narrows or disappears entirely,” the authors say. This calls for behavioral analytics; real-time validation systems for verifying payment details against authoritative sources before funds move; and phishing-resistant multifactor authentication (MFA). Finally, identity, device and context are continuously verified for every request to to enforce zero-trust identity and access management.

“These technologies must evolve as fast as the threats they’re designed to counter. A static defense posture is a losing one.”

The authors underscore the necessity of collaboration in facing the barrage of AI fraud. “No single organization can solve the problem of deepfake fraud alone. By sharing intelligence, best practices and technology innovation, organizations can strengthen collective defenses in ways no one can achieve independently.”

The future is layered 

Evidence continues to build that deepfake detection is transitioning from a standalone identity verification capability into a continuous trust layer spanning onboarding, authentication, communications, payments and fraud prevention.

The identity ecosystem is labeled as such because of how it all connects. Digitization is as much a story of convergence as it is one of plenty: more and more things happen online, and one searches for simpler ways to govern all of the connected parts. Meanwhile, fraud attacks once confined to skilled hackers targeting high-stakes financial and security institutions can now be engineered at scale, enabled by AI technology that also powers biometric injection attacks as a primary vector for deepfakes.

The ecosystem, then, presents itself as a cacophony in need of structure – or orchestration. The term, which connotes both arrangement of the component pieces and active guidance in how they interact, is a buzzword, but also an apt metaphor for what the modern identity stack needs. The soldier at the gates is no longer a lone giant stationed by the front door, but a gamesmaster meeting the shifting threat with nimble hands and a diverse arsenal of defenses, always on the lookout for alerts.

Should that test a general tolerance for whimsy, the Biometric Update 2026 Deepfake Detection Market Report, co-authored with Goode Intelligence, gives it to you straight: “defense is increasingly delivered as part of a layered identity security architecture rather than as a standalone product. Many vendors combine deepfake detection with presentation attack detection (PAD), injection attack detection (IAD), liveness detection and broader fraud prevention capabilities. While the technologies and target markets differ, the common objective is to help organizations detect and prevent AI-enabled identity fraud.”

The threat is expanding across digital pipelines and getting more efficient. Defenses must do the same: specific security elements working in concert to provide unbroken, ongoing protection against AI-driven fraud.

Article Topics

Accenture  |  AI fraud  |  biometrics  |  continuous verification  |  deepfake detection  |  deepfakes  |  IngenID  |  injection attack detection  |  JPMorgan  |  Resemble AI