Transparency note on Marvin's AI

This transparency note outlines how Marvin's AI works. We'll also explain its capabilities, limitations, and some best practices to maximize performance. We've included details of how you can make decisions to adapt system performance and behavior to your needs.

Marvin's AI makes use of pre-trained large language models, which have embeddings via Azure OpenAI Service.
​

We do not build or train our own models, and therefore do not have access to internal
model architecture, weights, or training datasets.

We make use of third-party transparency documentation regarding model architecture, usage constraints, and known limitations.
​

All communication between you, Marvin and the LLM models is encrypted. Your data may be used to train the LLM models. If you prefer that it is not used, you can opt out while signing up for Marvin.
​

The LLM processes the data according to its contractual agreements and privacy policies, while Marvin ensures compliance with industry standards.
​

While we do not control the foundational model, we ensure transparency during application of these AI models.

We ask you to check the accuracy and veracity of the data at each step. We also recommend that you verify all details before you communicate the information or take decisions based on it. This includes but is not limited to the following:

Moreover, all AI-generated content is labeled (e.g., “AI-generated insight” or “Suggested by AI”).
​

We have also included the option to edit AI outputs in research reports, discussion guides, transcripts, notes, AI analysis, etc., to make sure you have control over the data you use and communicate to others.

Our help center also includes various best practices to help you maximize efficiency while using Marvin's AI. You can learn more about this in each of these articles:

Your employer fully owns all the data you upload to Marvin.

You also have complete control over how you use Marvin’s AI. You can choose how to share data within and outside your team, and whether it should be included in Ask AI answers.

You can check the source of information and edit results before you share with others. We also give you the option to mask personally identifiable information (PII).

To make sure that Marvin works smoothly, we run tests and monitor performance regularly. Any of your data that may appear while checking the health of the system is pseudonymized, and all PII is redacted.