OpenAI has launched ChatGPT for Financial Services, combining GPT-6 Astra with licensed financial datasets, firm data connections and tools for modeling, research and client materials. Developed with Morgan Stanley and Evercore, the product shows OpenAI moving beyond the chatbot toward the full financial analyst workflow.
ChatGPT for Financial Services, announced on September 10, is a separate finance-focused plan built on ChatGPT Enterprise and delivered through the company’s ChatGPT Work environment. The significance is not simply that bankers now have another specialized chatbot. OpenAI is attempting to remove several of the hand-offs that normally sit between financial research, data terminals, spreadsheets and presentation software.
Financial data comes built into ChatGPT
The clearest change is the data layer. ChatGPT for Financial Services includes licensed datasets from providers including Daloopa, PitchBook, LSEG News and Crunchbase, covering areas such as company fundamentals, financial statements, earnings transcripts and private-market information. OpenAI says these included datasets can be used without setting up separate connectors or negotiating another contract with the provider.
The data is indexed and hosted on OpenAI’s infrastructure. That allows ChatGPT to retrieve figures and supporting passages directly and provide granular citations so users can trace numbers and claims back to underlying evidence. For firms that already subscribe to other information services, OpenAI is also working with providers including S&P Capital IQ, Dow Jones Factiva and Moody’s on entitlement integrations that can recognize an institution’s existing access. Its broader connector ecosystem includes more than 50 services, including FactSet, Preqin, Datasite, Box and Intapp.
The distinction matters. OpenAI is not trying only to make a model better at answering finance questions. It is trying to reduce the friction between the model and the licensed data professional analysts already depend on.
GPT-6 Astra moves from research into models and pitchbooks
OpenAI is pairing that data access with GPT-6 Astra. The company says the model is designed to retrieve information across multiple sources, interpret tables and footnotes, follow figures across reporting periods and perform financial reasoning. The resulting analysis can then be turned into editable spreadsheets, documents, charts and presentations.
Administrators can publish their firm’s own Excel, Word and PowerPoint templates inside the system, along with internal style conventions. In practical terms, the same workflow can move from earnings research or an LBO analysis to a valuation model, research note or pitchbook without requiring the user to rebuild the work in another environment. That is where the product becomes strategically more interesting than a vertical version of ChatGPT. OpenAI is positioning AI as the layer that sits between financial information and the final work product.
Compliance controls are part of the product, not an afterthought
That ambition also creates a different standard for security. ChatGPT for Financial Services inherits enterprise controls including SAML single sign-on, SCIM provisioning and role-based access controls. OpenAI says business data is not used to train its models by default, is encrypted in transit and at rest, and can be subject to configurable retention policies. Compliance teams can export supported workspace logs for audit and investigation workflows. Institutions can also create multiple workspaces to enforce information barriers and control which roles can access specific financial data sources, skills and app actions.
Those controls are particularly relevant in environments handling material non-public information, client confidentiality and restricted deal teams.
Integrated data does not eliminate verification
OpenAI’s own terms make clear that the system is not intended to replace professional judgment. ChatGPT for Financial Services does not constitute financial or investment advice, and OpenAI warns that both data and generated output can be inaccurate, incomplete, delayed or outdated. Users are instructed to verify important information, calculations and prices independently.
Licensing restrictions also remain attached to partner datasets. Depending on the provider, firms may face limits on bulk exports, redistribution, storage or use of data for model training. In other words, bringing licensed data into ChatGPT does not make the underlying contractual obligations disappear.
OpenAI is entering an increasingly crowded finance AI stack
OpenAI is also not entering an empty market. Anthropic has spent more than a year expanding Claude for Financial Services, adding financial-data connectors, Excel integration and finance-specific skills. In May 2026 it introduced ten agent templates covering tasks including pitchbook creation, financial modeling, earnings review, KYC screening, valuation review and month-end closing.
That competition highlights the larger shift underway. The next battle in enterprise AI may be less about which model produces the best standalone answer and more about which platform becomes embedded deeply enough in a regulated professional workflow that users no longer need to leave it. For OpenAI, ChatGPT for Financial Services is an attempt to make that workflow start with data, continue through analysis and end with the finished model, memo or pitchbook.