Cuvo’s stand on Open AI/LLMs

1. Cuvo’s Large Language Model (LLM) will not be enabled if the chatbot is not used.

The Cuvo’s feedback agent offers two-way communication for only human responses.

The Cuvo LLM will not generate any automated response for the feedback agent.

The chatbot can be disabled using a configuration setting that ensures the deactivation of LLM.

2. Data Send to Open AI.

The data that is being sent to Open AI includes the following.

• All two-way chatbot communications from a user are sent to Open AI.  All of our machine learning training material referred to as Knowledge Base are sent to Open AI.  This includes support tickets, technical documentation, or any other materials containing product intelligence that provide contextual relevance to the chatbot conversations.

3. The security and privacy measures for data sent to Open AI.

The security and privacy control on the data being sent to Open AI includes the following.

• The OpenAI complies with GDPR and CCPA. They can execute a Data Processing Agreement if your organization or use case requires it.
• The OpenAI API has been evaluated by a third-party security auditor and is SOC 2 Type 2 compliant.

https://openai.com/security

4. The on-premises LLM implementation options without data sent to Open AI.

The options that are available if we wish to implement on-premises LLM without the need for sending data to Open AI are as follows.

• Options for hosting:
  • Cuvo can host the LLM as part of our platform on our AWS
  • Cuvo (including the LLM) platform can be hosted on Block’s AWS
• LLM model options:
  • LLama 2 (7, 13, 70) billion models
  • Falcon 40 billion model

This will involve an additional 8-10 weeks to fine-tune the models.