OpenAI has launched a version of ChatGPT for banks and investment firms, with licensed market data built into the product. It is designed around the work financial analysts already do, including valuations, leveraged buyout models, buyer screening, earnings analysis, and pitchbooks.
The product was announced on Thursday. It runs on GPT-6 Astra and was developed with Morgan Stanley and Evercore. OpenAI is selling it to eligible financial institutions, although it has not published a price.
The more interesting part of the announcement is the list of data partners. It includes Daloopa, PitchBook, S&P Capital IQ, LSEG, MSCI, Moody’s, Dow Jones Factiva, Preqin, Intapp, Datasite and Box, with Reuters available through LSEG.
These are services that banks already pay for, and getting permission to use their data inside another company’s product can take time and involve complicated licensing agreements.
In this case, the data deals may be more important than the model itself. OpenAI is not just selling access to an AI system. It is putting a layer of software around information that financial firms already rely on.
OpenAI has also published a benchmark, although the result needs some context. On OfficeQA Pro, which tests financial document analysis, GPT-6 Astra scores 69.9%, compared with 60.2% for the previous model.
That is a meaningful improvement, but it also means the model still gets the test wrong around three times out of ten. In finance, an error can end up in a valuation model or a client presentation, which makes that distinction important.
The enterprise security features cover many of the requirements financial institutions would expect. They include single sign-on, user provisioning, role-based access, no training on business data by default, encryption at rest and in transit, configurable data retention, and audit log exports. OpenAI has also built information barriers into separate workspaces, which is a practical way of dealing with the Chinese walls that investment banks use to keep sensitive information apart.
What is less clear is how the product fits into the European regulatory environment. The announcement does not mention EU data residency, a specific European hosting region, or the Digital Operational Resilience Act, known as DORA. The rules have governed ICT third-party arrangements for EU financial firms since January 2025.
For a European bank, that is not simply a procurement question. DORA requires financial entities to keep records of their ICT providers, set contractual terms around areas such as subcontracting and audits, maintain exit strategies and carry out resilience testing. We have reported that many European financial firms are still not ready for those requirements.
There is another issue further down the road. DORA gives European supervisory authorities the ability to designate certain technology providers as critical ICT third-party service providers. Those companies can then face direct EU oversight, including inspections and recommendations that financial institutions may have to act on.
No AI model provider has been designated under that mechanism yet. But a system that becomes deeply embedded in valuation, research and deal workflows across several major financial institutions could eventually attract that kind of attention.
The Morgan Stanley connection is also worth watching. The bank helped design a product intended to reduce the time analysts spend on research and modelling. At the same time, Morgan Stanley analysts have predicted that European banks could lose a fifth of their jobs to AI. The two developments point in the same direction: banks are preparing to use AI more deeply in the work that has traditionally required large teams of analysts.
The launch also comes as OpenAI is moving quickly to put GPT-6 Astra into different parts of the economy. The model was released on 3 September with OpenAI’s own warning about its cyber capabilities, and ENISA was given access this week to test it. Within days, the same model is being offered to investment firms with financial data connected to it.
We have also followed ChatGPT Work since its launch, while European brokers such as Scalable Capital have already connected client assets to external AI assistants.
The consumer version of this shift is already underway. OpenAI has been working to connect ChatGPT to personal bank accounts, where the main pitch is convenience, and the bigger question is what happens to the financial data once it enters an AI system.
The institutional version raises the same question at a different scale. Instead of a person connecting a current account, a bank could connect its research, financial models, and deal information to the same kind of assistant.
For European institutions, then, the question is not simply whether the tool works. They will also need to know where their data is stored, which subcontractors have access to it, whether the contract meets DORA requirements, and what happens when the model is eventually replaced. Those details are not in OpenAI’s announcement, but they will be in the procurement process.
