Security and Control

When using DeepFellow, you have full control over the models you choose to use, including their specifications, deployment on cloud providers or your own infrastructure, and associated costs. This flexibility comes with the responsibility of selecting models that align with your organization's requirements and risk tolerance.

Choosing a Model

DeepFellow is compatible with any open-source Large Language Model (LLM), but it's essential to carefully evaluate the implications of each choice. We recommend using models like Mistral, which has been declared GDPR-compliant and trained on legally-obtained materials. This can help minimize the risk of legal complications related to intellectual property in generated content.

Considerations

When selecting a model, consider the following:

• Data source: Ensure that the model's training data is obtained legally and aligns with your organization's data policies.
• Model type: Choose models that generate content suitable for your use case, such as text, audio, or video.
• Training: Consider using models fully trained by you on your own intellectual property to maintain control over the generated content.
• Regulatory compliance: Verify that the chosen model meets relevant regulatory requirements, such as GDPR compliance.