In a recent blog post, IBM Automation CTO Jerry Cuomo outlined a series of risks associated with the enterprise use of ChatGPT, emphasizing that several critical risk areas should be carefully considered before implementing the technology. Cuomo's blog post concluded that only non-sensitive data could be considered safe with ChatGPT.
Cuomo raised concerns over the lack of control and knowledge that businesses have once their data is fed into ChatGPT. He highlighted the potential legal ramifications, where unintentional data breaches could lead to the exposure of partner, client, or customer data in ChatGPT' s training data, potentially landing companies in trouble.
The blog post also pointed out the risks related to intellectual property, indicating that leaked data might result in companies violating open-source agreements. It was noted that any sensitive third-party or internal company information entered into ChatGPT becomes part of the chatbot's data model , raising the possibility of being shared with others asking relevant questions.
OpenAI's ChatGPT API has data sharing turned off by default, but critics have noted that conversations on the web version are saved by default. Users have the option to decline saving conversations, but this comes at the cost of forfeiting the ability to continue conversations from where They left off and using their data to improve the model. As of now, there isn't an option to retain a conversation history without consenting to data sharing.
