Agentic fraud, where criminals use AI agents to build fake identities and hit hundreds of companies at once, is the threat behind Incode’s new TrustGraph. On October 14, 2026, Incode Technologies launched what it calls a privacy-preserving trust network that connects fraud and trust signals across institutions without pooling, exchanging or exposing customer data. In the company’s early access programme, institutions using it detected up to 53% more fraud than with their existing models alone.
Why Incode says agentic fraud needs a network
According to Incode, fraudsters can now use AI agents to create synthetic identities, forge documents and retry after every rejection. Because each bank or marketplace sees only its own applications, one agent opening accounts at several banks in an afternoon can look like a string of unrelated applicants. Incode projects that agentic fraud will account for 60% of all fraud attempts by the end of 2026, which is the company’s own estimate.
Sharing fraud information would help, but the company says institutions rarely do it at scale. Traditional approaches can require pooling customer records, which can clash with privacy rules, create a single target for attackers and expose one institution’s customer data to another.
How TrustGraph works, according to the company
Incode says TrustGraph runs on patented peer-to-peer cryptographic technology. It is anchored to verified biometric identities, so it can recognise repeat patterns even when a bad actor uses a new name, document or device, and it returns a fraud signal in real time at onboarding, login or a high-risk transaction. It can also link seemingly unrelated applications that share a device or document details, and give a positive trust signal for customers verified elsewhere.
On privacy, the company says customer records stay with the institution that holds them. Secure multi-party computation compares one-time cryptographic material to find matching signals without giving institutions access to each other’s records. There is no central data lake or broker, and each institution applies its own rules, so TrustGraph is meant to be intelligence rather than a shared blacklist.
Incode says the technology comes from more than a decade of research by Incode and Identiq, which Incode acquired in June 2026. TrustGraph is available now as part of the Incode platform.
The other side: questions about agentic fraud tools
The 53% figure is “up to” and comes from Incode’s early access programme. The announcement does not say how many institutions took part, how long the test ran, or how false alarms were counted. The claim to be “the first” such network is also the company’s own. Interac’s comments support network intelligence in general but do not verify TrustGraph’s results.
There are also privacy and security questions. Because the system is anchored to biometric identities, readers may ask how biometric data is collected, protected and handled if someone disputes a match. Wrongly flagging a genuine customer is another risk. Incode says each institution makes its own decision, which helps, but the release does not describe how customers can challenge an outcome. The cryptography is described as designed to protect the identities of institutions asking for and giving signals, yet the release shares no independent audit. Supporters would say fraud is already networked and defences must be too; skeptics would say any shared system raises new privacy and governance questions.
You may also like our stories on how AI can outpace human cryptography research and IBM’s 2026 data breach report.
My view on agentic fraud
As a writer, I feel the idea of sharing signals instead of data is one of the smarter responses to a fast-changing problem. Fraud rings do not respect company boundaries, so defences that stop at them will always be a step behind. Still, I would want independent test results, clear rules for biometric data and a way for people to contest mistakes before calling it a breakthrough. It could be a useful tool, and readers can decide how much to trust a vendor’s own numbers.
FAQ: Incode TrustGraph and agentic fraud
What is Incode TrustGraph?
It is a privacy-preserving trust network that lets participating institutions share fraud and trust signals through peer-to-peer cryptography, without pooling or exposing customer records.
What does “up to 53% more fraud” mean?
In Incode’s early access programme, institutions using TrustGraph detected up to 53% more fraud than with their existing models alone. It is the company’s figure, and details of the test were not shared.
Is TrustGraph available now?
Yes. According to Incode, it is available as part of the Incode platform and works alongside the identity verification and authentication processes institutions already run.