You are responsible for ensuring when and under what circumstances personally identifiable information can be shared with third parties.
Some large language models (LLMs), like the OpenAI models, can only be accessed via the LLM provider’s API. This means that in Crystal Chat when you select an API-based AI model, such as OpenAI’s GPT 4.1, OpenAI as the API provider receives your prompt. Despite legal terms that may state that prompt data is not used to train the provider’s LLMs, the data is still collected, subject to the third party’s terms of collection, processing, sharing and use and strength of their information security controls.
PII Protection mode provides you with enhanced (but not guaranteed) protections that certain types of PII are not shared with third parties. Please review the types of PII identifiable and supported in Crystal Chat [here] [link to What types of personally identifiable information (PII) are identifiable?]
