Artificial Intelligence (AI) at Datasite

Yes. AI is a broad term for functionality designed to make human-like decisions based on data and predefined rules. For instance, one of our AI tools analyzes various data points within a document to suggest how the document should be categorized, a task that would require specific instructions in regular computing.

AI takes many different forms, including generative AI. Generative AI has the unique capability to create new content, thereby expanding beyond the limitations of traditional AI. For instance, we have a generative AI tool that rewrites company descriptions to improve search results in Intelligence, our buyer finder app. While both AI and generative AI are integral to operations at Datasite, they differ in their levels of autonomy and creativity. 

Your deal information is not used to train external AI models. Datasite’s AI capabilities are deployed and managed internally, guaranteeing the privacy and security of your information. 

Your deal information is kept secure at Datasite through a comprehensive system of controls that prioritizes data privacy and security, including:  

• Deployment of all AI tools and models within Datasite systems, so that no data is provided to external systems or parties 
• Anonymizing training attributes thoroughly by transformed coding that cannot be reassembled, thereby maintaining the utmost privacy 
• Limiting interaction details to specific instances and using reinforcement data solely for the project from which it was gathered, all while ensuring no attributable data is shared outside of a user's project. 

As a customer, you retain all Intellectual Property Rights in the materials you provide to us. This includes all documents, materials, and information that you or any of your users provide to Datasite in conjunction with our services. It is your responsibility to ensure the accuracy, quality, legality, integrity, and appropriateness of all these materials.

Customer documents and materials are owned by the customer. Datasite owns the technology and is licensed by its customers to use anonymized and aggregated data. The anonymized and aggregated data does not include deal information but is a set of numerical representations for different data "attributes" derived from generic customer material characteristics. These numerical representations cannot be reassembled, thereby maintaining the utmost security. 

Project documents and materials are deleted. The residual anonymized and aggregated data is designed to be untraceable back to its original source, ensuring utmost security and confidentiality. The residual data is not disclosed to any third parties at any point in the process and is used only for product development. All the software used during this process is deployed and managed within the Datasite ecosystem and follows the same high security and compliance standards as the rest of the Datasite platform.  

Client documents and materials on the Datasite platform are kept for the duration of the client’s project or subscription and then deleted within 30 days of termination or earlier if requested by the customer. If needed, Datasite can provide certification of the destruction of all documents and materials.

Yes, Datasite uses reasonable and appropriate safeguards in its design and operation of AI to meet industry practices and evolving applicable legal and regulatory obligations for data security and privacy. Datasite’s AI is not used for marketing purposes or to make automated decision-making. Datasite continues to monitor the legal and regulatory landscape such as the upcoming European Union AI Act.

Datasite currently uses a mix of GPT 3.5, 4.0, open-source large language models (LLMs), and internal proprietary models. We continually test different models to see what works best and update when warranted. 

The models we use are not trained on our customer’s deal information but are a set of parameters learned from external data sets. Furthermore, these models are deployed privately by Datasite. That means that all data and information processed by these models are securely confined within our system and are not shared externally, nor used for training public/open-source models. This approach ensures your data remains private and secure within the confines of our environment.

Datasite uses a multi-layered approach to testing AI models and features, including: 

• Testing and validating datasets used to train our proprietary AI models. 
• Automated quality assurance (QA) testing of libraries and services used by AI features. 
• User acceptance testing by subject matter experts. 
  

All data used to train Datasite’s AI is anonymized and aggregated to remove all deal information (e.g., numerical values, addresses, locations, and personal data) and thereafter is used solely to improve our products and services. Any reinforcement data generated based user interaction is used solely for contextual reinforcement for the project from which it was gathered.

For further inquiries or additional assistance, please contact [email protected].