OpenAI packages ChatGPT for finance with licensed data and firm-controlled templates
OpenAI has launched ChatGPT for Financial Services for investment banking and equity research. Reuters and VentureBeat report that it combines GPT-6 Astra with built-in licensed data, granular citations, firm-controlled templates and governance tools. Morgan Stanley and Evercore served as design partners; access is limited to eligible financial institutions, and the announcement provides no public price.
Artificial Intelligence··Midday
A workspace dedicated to finance
Reuters reports that OpenAI is starting the new product with investment banking and equity research. ChatGPT for Financial Services puts GPT-6 Astra into workflows for research, financial modeling and preparation of client materials. Morgan Stanley and Evercore helped shape the initial use cases as design partners. The announcement describes a deployment distinct from the consumer product: access is open to eligible financial institutions, and no public price is provided.[1]
Data moves inside the product
VentureBeat says the package brings licensed financial data into the workspace alongside existing connectors and adds granular citations to claims in responses. Users can move from source retrieval to valuation models, research notes and presentations. The report also marks an important limit on coverage: bundling LSEG News does not mean that every LSEG dataset is automatically licensed. Each institution's access still depends on the entitlements it purchases.[2]
The template and governance layer
Both reports show that the product is more than a chat interface. Administrators can publish firm templates and style guides for Excel, Word and PowerPoint, allowing employees to turn research into outputs defined by the institution. Reuters relays the company's statement that customer data is not used for model training by default. Independent field results on accuracy, entitlement management and time saved in production workflows have not yet been published, however.[1], [2]
