Enterprise data protection combines organizational rules and technical safeguards for confidential business information used with dialogue agents and other language-model services. section 3.11
Common measures include:
- Restricting or banning employee submissions of confidential information. section 3.11
- Using providers or deployments that limit prompt storage, logging, and third-party sharing. section 3.11
- Choosing custom deployments or private infrastructure when additional protection justifies the cost. section 3.11
Corporate responses to confidential prompts
After employees submitted proprietary code and meeting notes to ChatGPT, Samsung limited prompt sizes. JPMorgan Chase later introduced an approved internal language-model wrapper after banning employee use of ChatGPT. section 3.11
Protection is not absolute: software bugs can expose chat histories or personal information, and submitted content may be used for model improvement without informed consent. section 3.11
