OpenAI announced that its popular AI system, ChatGPT, can now access users' bank account data in a read-only mode after its latest update. The capability is limited to ChatGPT Pro subscribers in the United States, and users must opt in to grant access.
The access is strictly read-only: the model cannot perform transfers, change account settings, or initiate any financial transactions. However, by seeing transaction histories — including timestamps, amounts, merchants and purchased items — ChatGPT can provide more detailed, personalized financial advice.
OpenAI frames the feature as an accuracy and utility improvement: the company says roughly 900 million people use ChatGPT weekly, and about 200 million users ask finance-related questions each month. OpenAI also claims that users' financial data will be kept more secure within ChatGPT than many users keep it themselves.
The company adds that collected data will be used, in anonymized form, to further train its AI models. Security professionals quoted by Gizmodo cautioned that linking bank data to an AI account still carries risks: if an attacker compromises a user's ChatGPT account, the detailed transaction records could be used to craft highly convincing phishing attacks.
The launch of this feature occurs as OpenAI explores a possible public listing and seeks revenue-generating capabilities for its flagship product. At the same time, the company is engaged in major litigation with Elon Musk, who was an early investor and alleges he was misled.
How these changes will play out in practice — including adoption, user responses and any new security incidents — will become clearer as the feature rolls out and more users opt in.
