Q: Does Jump use our firm or client data to train AI models?
A: No, Jump does not train AI models with your data. This is important because the SEC has cautioned against allowing sensitive client PII to be retained in AI models where they might leak out in unexpected ways.
Jump captures audio from your in-person and virtual meetings, then uses speech-to-text technology to transcribe these meetings. Jump then uses a type of generative AI technology called large language models (LLMs) to summarize raw meeting data into meeting notes, tasks, email drafts, and more.
LLMs are pre-trained on vast amounts of public text data collected from a wide range of sources, such as books, articles, websites, and forums. Jump's LLMs are pre-trained, which means that Jump does not require the use of your data for model training or retraining.
A good analogy here is a calculator. You perform your calculation in a calculator, and when you are done, the calculator clears, forgetting what just happened. Similarly, Jump's AI models provide you with summaries, but the model itself clears immediately and doesn't retain any of your input data.
Q: What are common compliance considerations with using Jump?
A: Jump’s goal is to ensure your firm not just maintains, but improves existing compliance practices. For example, using Jump helps to ensure high quality meeting documentation for every advisor, every meeting.
Here are a few common compliance considerations:
Consent and Disclosure: Best practice is to disclose to clients that you will be recording or transcribing the meeting for notetaking purposes. In some states this is legally required. Jump makes this easy in various ways, including both visual and audible disclosure options across virtual, in person, and phone meetings.
Protecting Client PII: Jump complies with SOC2 standards for security and data protection (including completing regular SOC2 Type II audits), and compiles with HIPAA protocols for handling sensitive data. Jump uses encryption and other security best pr
AI Model Safety: Jump does not train models with client or firm data. To keep the "human in the loop," Jump is set up so that any AI outputs are reviewed and approved by the advisor or other users before such outputs are synced, saved, or sent anywhere.
Recording Meetings: To accommodate your firm's recording policies, Jump provides flexible options for data capture, including Recording mode (full video and audio files captured for each meeting) or Summary Only mode (outputs only with no video or audio media)
Data Retention: To accommodate your firm's data retention and books and records policies, Jump provides flexible controls over what data is retained or auto-deleted.
Q: What are Jump's security practices?
A: Jump complies with SOC2 standards, and uses modern security practices including end-to-end encryption for data in transit and at rest, multi-factor authentication, and more. We engage in regular security audits including SOC2 Type II, continuous monitoring of our controls via Vanta, third party penetration testing, and regular software updates.
Learn more about our security practices at the Jump Trust Center.
Q: Where does Jump store data?
A: Jump hosts data on servers located in the United States (Iowa).
Q: How long does Jump store my data?
A: It's your choice. Meeting data is retained in Jump until you decide to delete it, either manually or via automatic deletion.