JumpCloud AI Labs Experiment - Pre-release - FAQs

• How do I get access to the Labs Experiment (Pre-release)? Please see. /alpha For more information.
• How can I provide feedback? Or get help? Via the AI channel in the The JumpCloud Lounge
• Why is this not hosted on console.jumpcloud.com? Why is the UI different? This is a research experiment to evaluate whether we can get our model to provide accurate, reliable and cost effective results. If we can achieve this then you may end up seeing this on the platform.
• I've asked a question but didn't get a very good answer. Why? We're working hard to make this as good as it can be. However, questions about the JumpCloud directory. Things like LDAP groups, SSO applications aren't supported yet. For the best results try asking a question about system insights. A full list of the data that is supported is at the bottom of the home page. Lastly, be more specific about the data you're after. For example when searching for admins we'll default to system administrators. If you're after console admins then include "console admins" in your query. Likewise, if for users we'll default to console users. For system users add "system users" to your query.
• Is my data used for training? No. No customer data is used to train, or given to 3rd party models. However, the questions that you submit and the SQL generated from them, are used to train models. Any identifiable information such as; names, IP addresses, hostnames etc are all manually reviewed and removed before they're added to the training set.
• Is my data safe? Yes, your data does not leave the existing JumpCloud production environment. Nor is it mirrored to a new service.
• How are you safeguarding against malicious queries? Or LLM injection? All questions asked and the returned SQL are sanitized. We have trained our model to evaluate all SQL generated to ensure no malicious queries are executed. However, if you find something like a potential security issue. Or would like more information reach out to us via The JumpCloud Lounge.
• What model is being used? We're evaluating many models to see what works best. However, the primary model used to answer questions is currently based on an open-source model trained on SQL. Which we've fine-turned alongside a Retrieval-Augmented Generation (RAG) to generate the SQL queries.
• Will you notify me if anything changes? Yes, we will update all Labs (Pre-release) participants by email and in our slack channel.
• Can I ask questions across multiple organizations? No, not at the moment. If you have an MSP account you can select each organizations from the drop down. However, you cannot query them in aggregate.