Anthropic announced Personal Finance in Claude on November 5, rolling the feature out in beta this week to subscribers on the Pro, Max and Ultra plans in the United States. Users link the bank and card accounts they choose through Plaid, the service that connects applications to banks, and then ask Claude about their balances, spending and bills in plain language. Reuters noted that the launch arrives ahead of Anthropic's anticipated initial public offering, and it is one of the clearest signs yet that the major chatbots intend to become the place where people manage everyday decisions, not only where they draft text.
The access Claude receives is deliberately narrow, and Anthropic publishes the list. For each connected account, Claude can see the bank name, the account name and type, the last four digits of the account number, balances and credit limits, and up to two years of transactions, and users can link up to 16 banks. Users sign in to their own bank through Plaid, so Claude never asks for or sees a bank password. The access is read-only: Claude cannot move, manage or invest money, and Anthropic says the feature is not designed to give investment advice. The company also says Personal Finance carries no ads or sponsored recommendations, offers no credit cards, loans or savings accounts of its own, and does not use conversations in the Personal Finance hub to train its models.
That read-only line is the whole product, and builders should study it. The banks drew the same line for outside agents earlier this autumn, allowing them to look at money but not to move it, because reading data and taking action carry different levels of risk and liability. By building on Plaid's approved, revocable connections instead of asking for credentials, Anthropic gets a feature users can plausibly trust, and it gets it without becoming a bank. Anthropic is also candid about the remaining risk: Claude can make mistakes, so users should ask which transactions sit behind any figure and check important numbers with their bank. A finance answer that cannot show its workings is not ready for real money, even when the underlying data is accurate.
The second story of the day shows the same pattern on the media side, where capability is arriving as a cheaper production service rather than a research preview. Google announced on November 5 that Gemini Omni 1.2 Flash is generally available across the Gemini API and Gemini for business, the first time a Gemini Omni model has reached general availability. The model creates and edits video natively at up to 15 seconds, up from 10 seconds in Omni 1.1, with durations selectable in one-second increments from 3 to 15 seconds. It adds multi-turn conversational editing and stronger reference following, so a creator can fix one detail in a follow-up prompt instead of regenerating the clip.
Google also cut the price by 30 percent from Omni 1.1, to 0.07 dollars per second for 720p video, and said every output carries an imperceptible SynthID watermark. The model is rolling out through Flow, Google Vids, Google Ads and the Gemini app, where 15-second videos are reaching Google AI Ultra subscribers first. Taken together, the two launches describe where applied AI is heading this winter: assistants that are trusted with more of a user's real data, under permissions that are explicit and reversible, and creative models that are priced and packaged for production work rather than for demonstrations. The winners will be the teams that design those permission and cost boundaries as carefully as they design the models themselves.