Looking for options to mitigate limitations of LLM while integrating with ODA.

Summary:

Looking for options to mitigate limitations of LLM while integrating with ODA.

Content (please ensure you mask any confidential information):

Below are the generally found limitations of LLMs. As the Oracle is providing this capability of integration LLMs with digital assistant, what are the suggestions or options provided to mitigate these limitations? 

Lack of Transparency

Potential Bias

Data Privacy

Lack of Control on response generated by LLM

Hallucination

Prompt Injection

Version (include the version you are using, if applicable):

Code Snippet (add any code snippets that support your topic, if applicable):