1. What is Claude Personal Finance?
Anthropic released Claude Personal Finance on Thursday, a beta feature that lets paying U.S. subscribers connect bank and credit card accounts and ask Claude questions about their money. Within hours, crypto investor Anthony Pompliano publicly accused the company of trying to extract proprietary details from his own startup, Silvia, months before launch. Anthropic has not been quoted responding to that allegation, though it has described how the new feature handles data.
The tool is aimed at everyday questions: how much was spent this month, which bills are coming due, and how much remains before the next paycheck. It sits inside Claude as a finance hub, and Anthropic has positioned it as the first step in a broader effort rather than a finished product.
The launch matters for two reasons. It moves a general-purpose AI assistant into one of the most sensitive categories of personal data, and it sets up a public dispute between a large AI developer and a founder who says the product duplicates his own.
2. How does the feature work and who can use it?
Personal Finance is rolling out in beta to subscribers in the United States on the Pro, Max and Ultra plans. Users link their accounts through Plaid, the data-connection service that handles the bank login step so that Claude itself never receives the user's banking password.
Once accounts are connected, Claude can summarize spending, flag duplicate charges or unexpected fees, explain how interest is being charged, and produce charts or calendars of upcoming bills. Anthropic's own illustration is a user who first asks where money went during the month and then asks what is left before payday.
Placing the feature inside the existing subscription also means no separate fee is attached to it, so the finance tool functions as an added reason to hold a paid plan rather than as a standalone product with its own price.
The beta is limited to the three paid tiers and to American customers. Anthropic has said it is exploring partnerships with licensed professionals and regulated financial firms as a next stage, which suggests the company sees the current version as an information tool rather than a regulated advisory service.
3. What data can Claude see, and what can it not do?
Anthropic stated on X that access is read-only. Claude cannot move money, manage accounts or make investments on a user's behalf. It also said that, like the rest of Claude, the Personal Finance hub contains no advertising and no sponsored recommendations.
The scope of visible information is specific. From accounts a user approves, Claude can see balances, credit limits, account names and types, the last four digits of account numbers, and up to 24 months of transaction history. Under Anthropic's documentation, conversations in the Personal Finance hub are not used to train its models.
The company also warns that balances may lag behind the bank's own records and that figures produced by the AI should be checked against official statements. That caution is common for assistants that summarize financial data, because a delayed or misread transaction can change the answer to a simple question such as how much is left to spend.
One point remains unclear from the public description: how Anthropic tracks users' financial information beyond model training, and how it may be used in other ways. Transaction histories are valuable to large data companies, which is why privacy questions tend to follow any consumer product that ingests them, even when the access is read-only and the company promises no ads.
4. Why does Pompliano say Anthropic copied his startup?
Pompliano runs Silvia, a startup that describes itself as a personal AI chief financial officer. Shortly after the announcement, he posted a lengthy complaint on X saying that Anthropic had effectively built a version of what his company offers, and that it had tried to learn how his product works before releasing its own.
The timing sharpens the complaint. Pompliano says the approaches began months earlier, so the product reached the market after a period in which he believes his company was being studied. Whether that sequence reflects ordinary competitive research, a partnership discussion that never matured, or something more deliberate is the question his post raises, and the facts available do not settle it.
His account begins in April. He alleges that people from Anthropic approached Silvia's engineers, citing interest in the startup's growth and its computing needs. In his telling, the aim was to understand how Silvia was built and why users found it popular. These are Pompliano's claims, and the public record so far contains no response from Anthropic to them.
Pompliano also describes a steak dinner with Anthropic representatives, which he says was reportedly billed to chief executive Dario Amodei. He says his engineers declined to share sensitive information, and that someone at Anthropic warned his team about the launch a few days before it happened.
5. Why is financial data such a sensitive category for AI assistants?
Money data differs from most information users share with chatbots. A conversation about a recipe reveals little, whereas a transaction list shows where a person lives, shops, travels and seeks medical care, and when income arrives. Assistants that handle this material therefore face higher expectations on consent, retention and security than general chat products.
For that reason, the details Anthropic has published carry weight: read-only access, a defined list of visible fields, a 24-month window and a statement that hub conversations are excluded from training. Each limits what could go wrong, and each is a commitment that users and regulators can later hold the company to.
Linking through Plaid adds another layer. The service sits between the bank and the application, which means the assistant receives approved data rather than credentials. Users still decide which accounts to approve, and the visible scope covers only accounts they choose to connect.
6. What is the dispute over the technology behind Silvia?
At the core of Pompliano's argument is his description of Anthropic's launch as a Claude wrapper presented as a personal finance product, meaning a thin layer over an existing model instead of a dedicated system. He contrasts that with Silvia, which he says relies on its own models and hardware.
He claims that approach cuts Silvia's computing costs by about 97% compared with running the same work on frontier models. He did not provide the details in the material available, though the figure implies a smaller model tuned for the task and run locally, rather than a call to a large general model for every query.
The claim, if accurate, would explain why Pompliano considers Anthropic's interest in his engineering meaningful. A company running a premium assistant for subscribers pays for every question in computing time, so a method that answers financial questions at a fraction of that cost has commercial value. It is also a claim that outsiders cannot check, and it should be read as the founder's own characterization.
7. How does this fit with other AI finance tools?
Anthropic is not the first AI developer to connect an assistant to bank data. OpenAI introduced a comparable feature for ChatGPT in May, which also uses Plaid and is read-only. Taken together, the two launches show the large model makers converging on personal finance as a consumer use case that justifies a subscription.
For startups building specialized finance assistants, that convergence is a competitive problem. A general assistant already has a large user base, a billing relationship and brand recognition, and can add a finance feature at little marginal cost. Smaller companies usually respond by pointing to focus, lower prices or better accuracy, which is the line Pompliano appears to be taking with his cost claim.
At the same time, general-purpose agents can perform similar tasks when given the right skills and permissions, so the boundary between a dedicated product and a configured assistant is thin. That ambiguity is part of why accusations of copying are hard to resolve: connecting accounts, categorizing transactions and answering questions about spending are widely used approaches, and no single company owns them.
8. What should users consider before linking accounts?
Read-only access limits the risk, because the assistant cannot initiate payments or transfers. It does not remove every concern. The information visible to Claude, including balances, credit limits and two years of transactions, can reveal habits, income and obligations, so users should weigh what they are comfortable sharing with any service.
Anthropic's statements address several of the most common worries: no advertising, no sponsored recommendations and no training on hub conversations. Users who want certainty can consult the company's documentation on retention and deletion, and can disconnect accounts through the connection service they used to link them.
Because balances can lag, the answers should be treated as summaries rather than authoritative records. A user planning around a tight payday should confirm the figure in the bank's own app before acting on it.
9. What happens next?
Anthropic describes Personal Finance as a first step and says it is exploring partnerships with licensed professionals and regulated financial firms. The beta stays restricted to U.S. customers on the three paid plans for now, and no date has been given for wider availability.
The dispute with Pompliano is still at the stage of public statements. Whether it leads to a formal response from Anthropic, a legal claim or simply a debate about competitive practices is not yet known. What is already clear is that the market for AI-driven personal finance now includes two of the largest AI developers and a growing set of startups, and that data handling will shape which products users trust.