D&B RESPONSIBLE GEN AI COMMITMENTS
D&B will not use Customer prompts or queries for the purpose of training or tuning any Gen AI models. D&B is committed to responsible AI, as stated in the Privacy, Data Transparency, and AI Trust Centre located at AI Systems and Use at Dun & Bradstreet.
Permitted Machine Learning License
Customer may, in connection with the permitted uses set forth in this Order, use the Information to develop statistical and machine learning models (e.g., regression, decision trees), provided that: (i) such models are designed to perform specific tasks by identifying patterns within the Information and making predictions or decisions based on those patterns; (ii) the models include documentation sufficient to reasonably explain decision-making processes and output, in accordance with industry standards for model interpretability (e.g., feature importance, decision paths, audit trails); and (iii) the models do not serve as a substitute for, replicate, or eliminate the need for the Information. Customer must have separate written permission from D&B prior to using D&B provided Consumer Data, Contact Information, or other Personal Data under this Permitted Machine Learning License.
Augmentation Use License
Customer is not authorized to use Information or Variations for Augmentation purposes unless Customer has purchased the applicable D&B.AI Package with respect to that Information. If Customer has purchased a D&B.AI Package, then, solely in connection with the permitted uses set forth in this Order and subject to the Restrictions below, Customer may use the Information provided as part of the D&B.AI Package for Augmenting and Grounding the outputs of Customer’s instances of pre-trained Gen AI models and Agentic AI tools (including embedded Gen AI features within proprietary or third-party software, such as Salesforce Einstein™ or Microsoft Copilot™). Customer must obtain separate written permission from D&B before using any D&B-provided Consumer Data, Contact Information, or other Personal Data under this Augmentation Use License.
Additional Permission Requirement
Customer must have separate written permission from D&B prior to using D&B provided Consumer Data, Contact Information, or other Personal Data under the permitted machine learning license and/or augmentation use license, as applicable.
Restrictions
Notwithstanding the foregoing, Customer shall not, and shall not permit any third party to: a) use the Information or Variations to develop, train, fine-tune, reinforce learning in, adjust weights in, customize or otherwise modify, enhance, improve or optimize any Gen AI; b) use the Information or Variations in any publicly-available or public-facing Gen AI; or c) commercialize, or distribute any output or action that is based on or derived from the Information or Variations.
DEFINITIONS:
- “Agentic AI” means software systems, models or applications that employ Gen AI to autonomously perform specified tasks, make decisions, or take actions with limited or no direct human interaction based on predefined objectives, learned patterns, or contextual data inputs.
- “Augmenting” or “Augmentation” means inputting, referencing, or otherwise entering the Information at Runtime, and solely as Read-Only Context, for the limited purpose of enhancing, informing, contextualizing, constraining, or guiding outputs or actions (e.g. Retrieval Augmented Generation (RAG)). Augmentation Includes the use of embedded Gen AI features within proprietary or third-party software, such as Salesforce Einstein™ or Microsoft Copilot™).
- “Gen AI” means an artificial intelligence technology that exists now or may be developed in the future that (i) is trained on data; (ii) interacts with a person using text, audio, or visual communication; and (iii) generates non-scripted outputs similar to outputs created by a human, with limited or no human oversight.
- “Grounding” means the use of Information to align Gen AI outputs with verified and contextually relevant data, for the purpose of improving factual accuracy, reliability, and relevance of such outputs. Grounding does not include training or fine-tuning Large Language Models (LLMs).
- “Read-Only Context” means the model can access Information only at prompt or query time, and only for retrieval, without modifying itself or memorizing the data.
- “Runtime” means the period during which Information may be used solely for executing a prompt or query and not memorized or reused outside a single prompt or query.
- “Variations” means information based upon, developed from, or otherwise incorporating all or any part of the Information, and which retains or conveys all or any portion of the original meaning, insight, analytical value, content, or conclusions of the Information.