• IBM Automation’s Jerry Cuomo recently published a blog post outlining the potential risks associated with using ChatGPT for enterprise.
• These risks include unintentional data leakage, violations of open-source agreements, and the sharing of sensitive third-party or internal company information.
• OpenAI has data sharing turned off by default in their API, but users must opt out of saving conversations and having their data used to train the model on web version.

Risks Associated With Using ChatGPT For Enterprise

Chief technology officer of IBM Automation Jerry Cuomo recently published a blog post laying out what he claims are several risks associated with using ChatGPT for enterprise. Listing numerous areas of concern, these potential pitfalls businesses should consider before operating ChatGPT include:

Unintentional Data Leakage

Cuomo warns that once your data enters ChatGPT, you have no control or knowledge of how it is being used. This type of unintentional data leakage could also put businesses on the hook legally speaking if partner, customer or client data is exposed to the general public after being leaked into ChatGPT’s training data.

Intellectual Property Risks

There are also potential intellectual property risks associated with implementing ChatGPT into your business operations. Companies must be aware that any sensitive third-party or internal company information entered into ChatGPT will become part of the chatbot’s data model which may then be shared with others who ask relevant questions.

Open Source Violations

Leakage could also put businesses in violation of open-source agreements as well as potentially exposing confidential information to those outside the organization who shouldn’t have access to it.

Privacy Features

OpenAI has responded by claiming that “the data will not be shared with others who ask relevant questions” and directed Cointelegraph to existing documentation on their privacy features including turning off chat history on their web interface and apps. However, critics point out that conversations on the web version are saved by default and users must opt out both from having their conversations saved as well as from having their data used to train the model given there is no option currently available to retain conversations without agreeing to share said data.