OpenAI launches ChatGPT for Financial Services, built with Morgan Stanley and Evercore

OpenAI rolled out ChatGPT for Financial Services on Sept. 10, targeting tasks like LBO modeling, buyer screening, earnings analysis, and pitchbook creation. The product goes head to head with Anthropic’s Claude for Financial Services, which uses tool connectors to tap bank data providers.

Built on GPT-6 Astra and hosted datasets

ChatGPT for Financial Services is described as a tailored version of ChatGPT Work, designed with Morgan Stanley and Evercore. It runs on GPT-6 Astra and combines the model with premium data from Daloopa, PitchBook, LSEG News and Crunchbase.

OpenAI said it hosts and indexes that data on its own infrastructure rather than requiring banks to build separate connectors. The announcement also says citations trace figures back to the relevant source filings.

Anthropic uses MCP connectors and provider entitlements

Anthropic’s Claude for Financial Services, announced May 5, connects to FactSet, S&P Capital IQ, Morningstar, PitchBook, Daloopa and Moody’s through pre-built MCP connectors. It also includes 10 ready-to-run agent templates for pitchbooks, KYC screening and month-end close.

Anthropic’s approach relies on connecting out to each provider’s system, with the model using the bank’s existing entitlements rather than hosting datasets itself.

Rogo and Hebbia face pressure from both directions

OpenAI’s launch adds to competition AI native startups Rogo and Hebbia have faced since Claude for Financial Services shipped. Both startups convert written instructions into financial models, memos and pitch decks, using a bank’s paid data and internal files, and they run on Anthropic, OpenAI and Google models.

Rogo reported passing $50 million in annual recurring revenue, up from $15 million at the end of 2025, and said it is on track for roughly $100 million by early next year, citing an unnamed source in a Sept. 3 report from The Information.

The Information also reported that once agents are fed a bank’s own history, they behave “increasingly like your firm rather than like a generic Claude,” an argument both startups make for why deep integration with a specific firm’s data matters.

The core dispute is control over workflow and data

OpenAI’s decision to license and host some datasets directly suggests a different model of control than Anthropic’s connector-based design. OpenAI’s Thursday announcement says selected datasets from Daloopa, PitchBook and LSEG News are bundled, indexed and hosted on OpenAI infrastructure, while other providers use lighter sign-in integration that recognizes existing entitlements.

The source also reports that banks are building internal AI agents on top of the same underlying models, which could reduce demand for third-party workflows even if banks still lack “scoped use cases” for effective in-house tools, according to a quote attributed to Farsight CEO Samir Dutta.

Why it matters

The launch accelerates a shift from general-purpose AI to embedded financial workflows, with OpenAI and Anthropic competing on both model performance and data access architecture. For banks, the choice between hosted datasets and connector-based retrieval will shape how citations, retrieval control, and integration costs play out across research and deal work.