On April 9, 2026, the U.S. Patent and Trademark Office published five patent applications from Wells Fargo Bank, N.A. on the same day — and four of them carry the identical title, “Systems and methods for managing digital assets and digital asset transactions.” A published application is not an issued patent; it is a roughly 18-month-delayed look at where a company directed filing effort. When five land together, from an overlapping inventor team, the cluster itself is the thing to read, not any single document.

The hero of the group, US20260099480A1, describes the core operation: take a file of text strings, validate the information in it with a machine-learning model, then turn the file into a token recorded on a ledger. The abstract states it directly:

In response to determining the validity of the information in the text strings, tokenize the file by generating a digital asset corresponding to the file, the token includes a pointer to the file, and record the digital asset on a distributed ledger database/blockchain.— Systems and methods for managing digital assets and digital asset transactions, US20260099480A1

The token does not hold the file; it holds a pointer to it, with the validation handled by an ML model before tokenization. The classes on the record — G06F 16/2255, G06F 21/602 (data encryption), and G06F 40/284 — describe document processing and encryption rather than payments. This is disclosed as a way to represent a document as an on-chain asset, not as a coin or a settlement rail.

The same idea, applied four ways

What makes the week worth a column is that the team applied the tokenization pattern across distinct assets in parallel. US20260099571A1 tokenizes a machine-learning model: a first token for the model, a second token for the model's intellectual-property rights with bidirectional links between them, and a third token for the training dataset, all recorded on a distributed ledger. US20260099839A1 takes two versions of an ML model from two separate blockchain registries, records both on a third ledger, and selects a “champion model” between them — a model-governance step expressed as on-chain assets. And US20260100855A1 describes a “universal resolver” that reconciles provider-institution and user identifiers pulled from two different blockchain registries into one universal unique identifier recorded on a third ledger; that filing is the one carrying the blockchain-specific class H04L 9/50 together with H04L 9/0825.

A fifth application from the same day, US20260099850A1, sits adjacent: a token-management system on a mobile device that links, deactivates, reactivates, and deletes a payment token tied to an account. It is a payment-token filing (CPC G06Q 20/405, G06Q 20/3226) rather than a ledger filing, but it shares the week's vocabulary of tokens and lifecycle control.

What the cluster signals, read through the structure

The business reading follows from the assets being tokenized. The filings are not about cryptocurrency; they are about representing a bank's own operational objects — documents, ML models, the IP rights attached to those models, training data, and cross-institution identities — as tokens whose state lives on a distributed ledger. Three of the four share an inventor team (Anita Sukur, Ian T. Staley, and Ashmita Regmi), which indicates a single program rather than scattered experiments. The recurring elements across the cluster are validation before tokenization, pointers rather than payloads on-chain, and reconciliation across multiple blockchain registries via a resolver.

Put together, the disclosed direction is infrastructure for asset and model governance on a ledger: a way to register what a document or model is, prove which version is authoritative, attach rights to it, and resolve identities across institutional boundaries — with the heavy data kept off-chain and only the token and its pointer recorded. The emphasis on ML-model tokenization and a “champion model” selection step is notable for a bank, and points toward governance of machine-learning assets as a filing priority, expressed in on-chain terms.

The caveats apply with force. These are published applications, not grants; claims can narrow or fail before issuance, and a filing discloses intent to seek coverage, not a deployed system. Blockchain publication volume is thin across the sector, and in the April 7 to April 13, 2026 window the records indexed for the sector keyword set number 72 published applications, the large majority from unnamed assignees or carrying only incidental ledger mentions. Against that thin field, Wells Fargo is the single largest named filer of the week, with five applications — and four sharing one title. That concentration, from one team, in one week, is what the record actually supports: a bank documenting, in detail and in parallel, a way to put documents and machine-learning models on a distributed ledger as governed tokens.