Data FAQ for Pearl & Precision

Is Pearl user data used to train or improve any Gen AI/LLMs?

No. Pearl user data is not used to train or improve the models of Unified’s third-party artificial intelligence service partners, such as OpenAI. You can view OpenAI’s Data Processing Addendum here. In addition, under Open AI’s Zero Data Retention policy, Unified content such as prompts and responses, as well as classifier outputs, are excluded from OpenAI’s abuse monitoring logs.

When a user uses Pearl, what data is sent to third-party artificial intelligence service(s)?

Identifying information, including anonymized information, is not sent to a third-party artificial intelligence service. The elements to be charted and the text of the prior art are sent to the third-party artificial intelligence service.  All communications with the third-party artificial intelligence service are encrypted.

Is Pearl user data saved?

Yes in the industry-standard, encrypted Amazon Relational Database Service. In general, when a user uses Pearl to generate a chart, Unified Patents saves an anonymized user ID and a timestamp when a chart was generated. Additional information may be saved in the following circumstances:

1. If the user clicks the thumbs up or thumbs down button on a quote, PEARL saves the corresponding limitation, prior art, citations, and timestamp. This data is anonymous and it is not stored with or associated with any user identifier.  This data is used internally by the Unified team as feedback for future improvements (it is not used to train any model).

2. PEARL offers a workspace feature that allows users to save and restore their progress. If this feature is used, data such as target patents and prior art references will be saved for the purpose of restoring the session. However, this data is stored using only an anonymized user ID and is not directly tied to any identifiable user information.  Importantly, this feature is opt-in.
    
3. If the workspace feature is not used and the thumbs up or thumbs down buttons are not clicked, then no record is kept of which target patents or prior art references were used at all. In such cases, we only log general usage metrics associated with an anonymized user ID for internal tracking purposes (e.g., usage counts), and no specific content or context from the charting activity is retained.