May 18, 2026
Ti3 is here: A bigger graph for a fast moving fraud landscape
Today, we’re announcing Trust Index 3 (Ti3), the latest version of the machine learning model powering Plaid Protect. Ti3 expands how Protect detects fraud across the Plaid network with a dramatically larger fraud graph, deeper real-time relationship analysis, and new signals designed for modern attack patterns.
The result is a fundamentally different approach to fraud detection, one built on visibility across the network, not just within users from a single platform. Early testing results are promising: Ti3 catches up to 41% more fraud while maintaining the same false positive rate as previous models.
A larger graph built for coordinated fraud
At the core of Ti3 is a major expansion of Plaid’s fraud graph, a network map of how devices, accounts, identities, sessions, and institutions connect across the financial ecosystem.
By understanding how those entities relate to one another over time, Ti3 can identify coordinated fraud patterns that would otherwise remain invisible inside a single app or transaction.
Our graph has grown nearly 10x in just a matter of months, dramatically increasing the number of relationships we can analyze across devices, accounts, institutions, sessions, and identities.
Ti3 can also now traverse that graph up to nine hops deep in real time, essentially following multiple degrees of separation between devices, accounts, identities, and sessions to uncover coordinated fraud activity that would otherwise appear unrelated.
That means Ti3 can identify patterns like:
A device opening accounts across multiple platforms in rapid succession
Shared infrastructure linking seemingly unrelated identities
Repeated account linking and disconnection behavior across the network
IP locations that are geographically inconsistent with the addresses tied to connected bank accounts
Device and session activity that deviates from established account behavior patterns
Often, the difference between stopping one fraudulent event and uncovering an entire fraud ring comes down to visibility across the network.
New signals designed for emerging attack vectors
More actionable fraud attributes
Ti3 introduces new signals and attribute categories designed to address increasingly sophisticated fraud patterns, while making the underlying insights easier for fraud teams to understand and act on.
As attacks evolve, fraud teams need more than just accurate risk scores. They need clearer signals that explain why activity appears suspicious. With Ti3, we’ve expanded and refined the attributes powering Protect to surface more actionable intelligence across account behavior, device activity, identity patterns, and network relationships.
More accurate account age intelligence
Synthetic identity fraud and bust-out schemes often depend on making new accounts appear established. Ti3 combines bank account age, transaction history, and network connection patterns to better distinguish genuinely tenured accounts from recently created ones.
Enhanced device identification
Account takeover attacks have become increasingly sophisticated as fraudsters rotate devices and spoof fingerprints to evade detection. Ti3 introduces enhanced device identification powered directly through Plaid Link, enabling more precise identification that is significantly harder to bypass than traditional device fingerprinting approaches.
Built for the future of fraud detection
For over a decade, Plaid has served as infrastructure connecting thousands of financial applications and the banking system. Every linked bank account, authentication event, and device interaction contributes to a broader understanding of how fraud moves across the ecosystem.
Nearly one million people connect through Plaid every day.
That network visibility enables fraud intelligence that is difficult to replicate because it was built over years of real-world financial activity, not assembled after the fact.
But the advantage isn’t just the scale of the network. It’s the speed at which we improve it.
We treat fraud prevention as a race, not a finish line. Ti3 is the latest result of a continuous model enhancement strategy where we ingest, analyze, and deploy ecosystem-wide intelligence at a velocity the rest of the industry struggles to match.
And we’re continuing to invest in what comes next.
To learn more about Protect and Ti3, reach out to our team today.