1. Does Emma test tested for bias, discrimination, and hallucinations?
Emma doesn't test itself any models as we use 3rd party models and we evaluate these models against their Model Cards that describe sensitivity towards bias, discrimination and hallucinations.
The model cards can be found below:
Model | Model Card |
Anthropic - Claude | |
OpenAI - GPT | |
Google - Gemini |
2. Does Emma follow the principle of human oversight?
Human oversight is built into Emma Legal's product design: the platform supports legal review rather than replacing it. Every AI-generated finding is linked back to the underlying source document or even clause-level in the data room, so the legal professional verifies each output before being reported.
3. Who owns the Intellectual Property of AI-Generated outputs when using Emma?
The client of Emma will own all IP. Emma has following terms enforced from its subprocessors and passes this IP ownership over to its clients:
1. Google Gemini The customer owns the output. Google does not claim ownership of content generated by Gemini. → Gemini API Additional Terms: https://ai.google.dev/gemini-api/terms
2. Anthropic Claude The customer owns the output. Anthropic assigns all right, title and interest in Outputs to the customer, also via AWS Bedrock: https://aws.amazon.com/legal/bedrock/third-party-models/
3. OpenAI GPT The customer owns the output. OpenAI assigns all right, title and interest in Output to the customer, who also retains ownership of all Input.
OpenAI Business Terms: https://openai.com/policies/business-terms
Microsoft Product Terms: https://www.microsoft.com/licensing/terms/productoffering/MicrosoftAzure/EAEAS