1. INCORPORATION AND DEFINITIONS These Foundation Model Supplementary Terms constitute an integral part of the Master Services Agreement or Personal Use Terms of Service ("Agreement") between you and Dust. These terms govern your access and use of Foundation Model features, including Large Language Models ("LLMs") and content generation models provided by Foundation Model Providers ("FMPs"). All capitalized terms retain their meanings as defined in the Agreement.

2. SERVICE EVOLUTION AND MODIFICATIONS The Customer acknowledges that Foundation Model Provider services are subject to technological evolution and policy changes. Dust reserves the right to modify these FM Terms to reflect significant changes in FMP services or policies. The Customer explicitly acknowledges that Dust bears no liability for FMP-initiated changes and cannot guarantee that such changes will not impact Dust's Services.

3. DATA PROCESSING AND PRIVACY a. Customer Data Processing: You authorize Dust to transmit Customer Data (definitions here), including Customer Personal Data, to Foundation Model Providers in accordance with the Data Processing Agreement ("DPA") and service functionality requirements. DPA agreements with each FMP can be found in the Sub-Processors list.

   b. Model Training Restrictions: i. Dust and FMP (OpenAI, Anthropic, Mistral, Google, Fireworks.ai) are prohibited from using Customer Data for model training ii. Dust may not grant third parties rights to use Customer Data for model training iii. Your use of Dust's Services does not confer any rights to use Customer Data for training purposes

   c. Data Retention from FMP:

   i. FMP (OpenAI, Anthropic, Mistral, Google, Fireworks.ai) apply a “Zero Data Retention” policy, meaning that Customer Content will not be logged for human review and will not be saved to disk or retained by those providers. Certificates can be found in the Sub-Processors list.

   d. Model Provider Selection

   Customers have the ability to opt out of any FMP used for embedding (upon request) and for inference (directly through the user interface).

4. INPUT AND OUTPUT PROVISIONS a. Definitions:

• "Inputs" refers to content submitted for processing
• "Outputs" or "Foundation Model Outputs" refers to FMP-generated content
• Both Inputs and Outputs constitute User Content under this agreement

b. User Responsibilities: You maintain sole responsibility for: i. Input and Output usage compliance with applicable laws ii. Adherence to these FM Terms, the Agreement, and Acceptable Use Policy iii. Respecting intellectual property rights

1. PROHIBITED ACTIVITIES Users are expressly prohibited from: a. Developing competing foundation models or services b. Using Outputs to deceive or mislead c. Generating manipulative content (including spam and electoral campaign materials) d. Violating Dust's technical documentation or usage parameters

2. THIRD-PARTY COMPLIANCE Users must comply with all applicable Foundation Model Provider policies, including but not limited to OpenAI terms and policies, Anthropic terms and policies, Mistral AI terms and policies, Google Generative AI APIs Terms of Service.

3. WARRANTY DISCLAIMER a. Dust explicitly disclaims all warranties regarding: i. Results obtained through service use ii. Foundation Model Output accuracy

   b. Risk Acknowledgment: i. Service use and Output utilization occur at user's sole risk ii. Users must independently verify Output accuracy iii. Detailed or specific Output may contain inaccuracies iv. Outputs may not reflect post-training developments

   c. No oral or written information provided by Dust creates additional warranties beyond those expressly stated herein.