Plaid has released a series of updates to its Identity Verification (IDV) product, targeting fraud prevention across document verification, data source verification, and case management. The updates come as new account fraud emerged as the only major fraud category to increase in both victim count and dollar losses last year.
Rising new account fraud and AI-manipulated documents
According to a study by Javelin Strategy, new account fraud increased by 31% in victim count and 13% in dollar losses in 2025, the only major fraud typology to rise across both measures. Plaid said fraudsters are increasingly using generative AI tools to alter identification documents in ways that are difficult to detect visually or through manual review.
Document verification updates
Plaid's IDV integrity check now inspects image metadata to flag tampered identification documents, including images created using virtual camera software or signs of digital editing or AI generation. According to Plaid, this feature flagged more than 50 virtual camera attempts within a single month. The product also includes detection for screens or printed copies of identification documents, addressing cases where fraudsters present physical copies or photographs of stolen IDs during online account opening.
IDV now also requires both sides of an identification document to be submitted. If only the front of a document, such as a driver's licence, is uploaded, users are returned to the document selection step with feedback indicating what is missing. Plaid said this is intended to close a gap that has previously been used to bypass verification and to reduce manual outreach for organisations with stricter compliance requirements, such as financial institutions.
Data source verification updates
Plaid IDV now includes an additional email risk check using a large language model to cross-reference a user's name and date of birth against their submitted email address to identify mismatches. Rather than automatically failing a session, the system flags suspicious patterns as an additional risk signal. According to Plaid, the feature identified all fraudulent attempts during a live attack on a major customer within its first two weeks of use and is available through both the dashboard and Plaid's API.
The product also includes a typo detection and correction feature. If a user enters an incorrect date of birth, Social Security number, or name during account opening, they are given the option to correct the error rather than having the session fail automatically. Plaid said this is intended to improve pass rates for legitimate users who make unintentional data entry errors.
Case management updates
Plaid has also updated its case management tools, aimed at giving investigators more information within a single view. The IDV dashboard now includes a location map showing a session's document address, submitted address, IP address, and VPN-associated IP address, along with the distance between them and an associated risk level, intended to make discrepancies such as an unexpected country of origin more visible.
Plaid has also expanded the fraud labels available when marking a session as fraudulent within the dashboard. The expanded categories include user account and identity fraud (such as account takeover, stolen identity, synthetic identity, multiple accounts, and scam victim cases), bank account or payment method fraud (including bank account takeover, revoked bank connections, and card testing), transaction-related fraud (including unauthorised transactions, chargebacks, ACH returns, and disputes), and behavioural categories (including first-party fraud, missed payments, loan stacking, and money laundering). Plaid said the expanded labels are intended to support faster categorisation, clearer investigation records, and improved training data for future fraud detection.
Implementation and broader risk strategy
Plaid said the updates are self-serviceable and do not require additional integration work and are currently available. The company recommends organisations adopt a defense-in-depth approach to fraud prevention, combining data source verification, document verification, liveness checks, behavioural analytics, device fingerprinting, and network intelligence, tailored to their individual risk appetite.
