Ask BankingNewsAI

Ask a question about AI in banking. Every sentence links its source.

Records as of Oct 8, 2026 · 10 questions per account · answers take about 10 seconds because every citation is checked

Answers come only from our profiles of 120 US banks, 20 non-US banking groups, 20 fintechs and 49 AI vendors, 237 regulatory documents from 41 authorities, and our daily briefs. Every sentence links its source. If we don't have it, we say so.

Loading

1 · Finding recordsThe question is matched to the entities, documents and topics in our records: the bank pages, the regulation tracker, the vendor and fintech dossiers, the rankings and the daily briefs.
2 · Reading themThe answer is written from those records alone, one claim per sentence, each sentence tied to the record block it comes from. Numbers, dates and status words are copied, never computed.
3 · Checking every citationA verifier compares each sentence with the text it cites. A sentence that fails is rewritten once or removed. If the records do not cover the question, the answer says so and points to the nearest records.

Banks

How many JPMorgan employees have access to LLM Suite?

JPMorgan — LLM Suite access: ≈ 250,000 employees, about half daily users (as of Apr 29, 2026).

our record · American Banker, Oct 21, 2025

How many active users does Bank of America's Erica have?

BofA — Active Erica users: 24 million+ (as of Jun 30, 2026).

our record · The Motley Fool (Yahoo Finance), Jul 21, 2026

Regulation

Does SR 26-2 apply to generative AI?

No. SR 26-2 states that generative AI and agentic AI models are novel and rapidly evolving and are not within its scope. Banks are instead expected to govern them through broader risk-management and governance practices, and the Fed, OCC, and FDIC said they would issue a request for information on AI and model risk. (SR 26-2, status: In force.)

our record · Federal Reserve, Apr 17, 2026

What is the status of the EU AI Act, and who does it apply to?

Regulation (EU) 2024/1689 — status: In force; published Jul 12, 2024; effective Aug 1, 2024. Providers and deployers of AI systems placed on the market or used in the EU, including banks, lenders, insurers and fintechs, and non-EU firms whose AI outputs are used in the EU.

our record · EU AI Act, Jul 12, 2024

Vendors

How much does Microsoft 365 Copilot cost a bank?

Microsoft Copilot publishes a price for Microsoft 365 Copilot, plan "Microsoft 365 Copilot (enterprise)": $30.00 per user per month, paid yearly (annual subscription); a qualifying Microsoft 365 plan is required separately (as of Oct 5, 2026).

our record · Microsoft, Sep 24, 2026

Which of the largest US banks use Feedzai?

Among the 120 largest US banks, Feedzai's page names as customers: Citi (In production), SoFi (Selected).

our record · Feedzai, Oct 11, 2024

What does Ask BankingNewsAI answer from?

Only BankingNewsAI's own records: AI-strategy profiles of 120 US banks and 20 non-US banking groups, 20 fintechs and 49 AI vendors, 237 regulatory documents from 41 authorities, 5 side-by-side comparisons, the AI disclosure index and the daily brief's summaries. It has no other knowledge for this task.

How are the answers checked?

Each sentence is generated with a citation to a record block. Before anything is shown, a verifier checks that every number, date and status word in a sentence appears in the record it cites, that every organisation named is an entity of that record, and that uncited sentences carry no facts. A sentence that fails is regenerated once or removed; if nothing passes, the answer is withheld.

What does it not do?

It does not predict, advise, or say whether a bank is compliant; it describes what a document says. It does not rank banks or vendors unless a ranking record exists, and it says so when the records do not cover a question.

We keep your questions and answers so you can revisit them and so we learn what we can't answer yet. We don't sell or share them, and they're not used to train AI models. Full details in the privacy policy and terms. Questions and the retrieved records are sent to Anthropic to generate the answer; Anthropic does not train on API data.