What is ChatGPT for Financial Services?

ChatGPT for Financial Services is a tailored ChatGPT Work product, announced by OpenAI on September 10, 2026, that bundles premium financial datasets with GPT-6 Astra's reasoning so banking and research teams can produce research, financial models and client materials in one place. It was shaped by design partnerships with Morgan Stanley and Evercore, starts with investment banking and equity research, and is available to eligible financial institutions through OpenAI's sales team rather than self-serve.

What financial data is built into ChatGPT for Financial Services?

At launch it includes datasets from Daloopa, PitchBook, LSEG News and Crunchbase — covering earnings transcripts, financial statements, company fundamentals and private companies — indexed and hosted on OpenAI infrastructure, with no separate contracts or connectors to set up. Granular citations let bankers trace a figure or claim back to the specific table or passage. For firms with existing subscriptions, OpenAI is building shared sign-in and entitlement integrations with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva and Moody's, and its connector ecosystem covers 50+ tools including Datasite, Box, Preqin, FactSet and Intapp.

How does ChatGPT for Financial Services handle confidential data and compliance?

It builds on ChatGPT Enterprise controls: SAML SSO, SCIM provisioning and role-based access. Business data is not used to train OpenAI models by default, is encrypted at rest and in transit, and admins configure workspace retention. Compliance teams can export supported workspace logs through the OpenAI Compliance Platform, firms can enable or disable app read and write actions by role, and multiple workspaces can be created to enforce information barriers between teams — the standard MNPI safeguard in banking.

How does ChatGPT for Financial Services compare with Claude for Financial Services?

Both are vertical packages of a frontier model plus financial-data connectors and enterprise governance. OpenAI's September 2026 launch differentiates on hosting premium data itself (Daloopa, PitchBook, LSEG News, Crunchbase indexed on OpenAI infrastructure) with granular citations, on entitlement sign-in for Capital IQ, LSEG, MSCI, Factiva and Moody's, and on firm-template artifact generation for pitchbooks and models on GPT-6 Astra. Anthropic's offering, launched earlier and covered on andrew.ooo, leans on its agent skills and connector marketplace with Claude models. Choice in practice follows which model your firm has already cleared through security review.

Quick Answer

ChatGPT for Financial Services Explained (September 2026)

Published: September 11, 2026

The Short Answer

On September 10, 2026, OpenAI shipped ChatGPT for Financial Services: ChatGPT Work with premium market data indexed inside it, GPT-6 Astra doing the analysis, and bank-grade governance on top. It was designed with Morgan Stanley and Evercore, starts with investment banking and equity research, and is sold to eligible institutions rather than offered self-serve.

	Detail (as of September 11, 2026)
What	Tailored ChatGPT Work experience for financial institutions
Design partners	Morgan Stanley, Evercore
Initial focus	Investment banking, equity research
Built-in data (hosted by OpenAI)	Daloopa, PitchBook, LSEG News, Crunchbase — transcripts, statements, fundamentals, private companies
Entitlement integrations (in progress)	S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, Moody’s — sign in with ChatGPT, get the data you already pay for
Connectors	50+ including Datasite, Box, Preqin, FactSet, Intapp; tuned MCP connectors for the most-used finance providers
Model	GPT-6 Astra natively; newer models “out of the box” as released
Outputs	Research, financial models, pitchbooks and client materials in the firm’s templates; interactive charts
Governance	SAML SSO, SCIM, RBAC, no training on business data by default, encryption, retention controls, Compliance Platform log export, per-role app actions, multi-workspace information barriers
Availability	Eligible financial institutions via OpenAI sales / account team

Why OpenAI hosts the data itself

The launch post is candid that MCP connectors to data vendors were the pain point: “reliable access to data and high quality artifact creation proved to be the biggest pain points” for Morgan Stanley and Evercore teams. OpenAI’s fix is to index and host Daloopa, PitchBook, LSEG News and Crunchbase on its own infrastructure. Two consequences:

No contracts or connectors to negotiate before an analyst can start — the datasets are ready on day one.
Granular citations. Because OpenAI controls retrieval, it can highlight the exact table or passage behind a number. The worked example: a banker normalising a P&L can inspect the reconciliation and notes behind an adjusted EBITDA, see which costs were excluded, and decide how to treat it in a valuation.

OpenAI also says it will post-train its models to “find, interpret, and use this data like we know the best analysts can” — the datasets are training targets, not just retrieval sources.

For firms that already pay Capital IQ, LSEG, MSCI, Factiva or Moody’s, the model is different: providers recognise the user through ChatGPT sign-in and unlock the data they are entitled to. Those integrations are described as in progress at launch.

What GPT-6 Astra brings

OpenAI positions Astra as state of the art on the three capabilities the product needs: information retrieval across sources, financial reasoning (take inputs, run analysis, draw conclusions), and artifact generation (documents, spreadsheets, slides). Concretely the product can research across multiple sources, trace a figure across periods, interpret annotations in public filings, build interactive charts, and produce valuation models, research notes and pitchbooks in the firm’s format — with templates published centrally to the teams that use them.

GPT-6 Astra’s API list price is $10 per million input and $50 per million output tokens; inside ChatGPT for Financial Services it is bundled into the enterprise contract, so the cost question is seat pricing negotiated with OpenAI, not tokens.

Governance for MNPI

Banks care about one thing above the model: material non-public information. The controls listed:

SAML SSO, SCIM provisioning, role-based access — inherited from ChatGPT Enterprise.
Business data not used for training by default; encrypted at rest and in transit; admin-set retention.
Log export through the OpenAI Compliance Platform into existing audit and investigation workflows.
Skills and apps managed by role; supported app read/write actions can be enabled or disabled.
Multiple workspaces to enforce information barriers between, say, M&A advisory and research.

How it compares (September 2026)

	ChatGPT for Financial Services	Claude for Financial Services (Anthropic)	Gemini Enterprise (financial vertical)
Model	GPT-6 Astra ($10/$50 per MTok list)	Claude Opus 5 / Fable 5.1 family	Gemini 3.x
Data approach	Premium data hosted and indexed by OpenAI + entitlement sign-in	Connector/skills marketplace to vendors	Google-side connectors
Citations	Granular, to table/passage	Source-linked	Source-linked
Design partners	Morgan Stanley, Evercore	Announced with several banks and data vendors in 2026	—
Distribution	Sales-led, eligible institutions	Sales-led	Google Cloud

The differentiator OpenAI is betting on is hosted data with citations; Anthropic’s is a broader skills ecosystem. In practice the choice is usually decided by which vendor’s model and data flow the firm’s security and compliance teams have already approved.

Who it is for, and who it is not

For: investment banking and equity research teams at institutions that can go through OpenAI sales, particularly those without existing data connectors wired into their AI tooling.
Not yet: retail advisors, insurers, asset managers outside equity research — OpenAI says partners will guide “expansion into other financial services categories.” Developers can still build on the API; the launch post explicitly invites that.