We encourage pilot participants to submit questions, issues, feedback here. Your input will help us develop the documentation and plan for the broader rollout.
Q: What is PennChat?
A: As part of ISC's AI Hub, we are piloting PennChat, a secure AI portal that brings together multiple chat-based Large Language Models (including Claude and ChatGPT) in a single, customizable interface. It is designed for use by Penn faculty, students, and staff, with plans to expand to a campus-wide offering soon.
Q: How do I access PennChat?
A: To use PennChat, you must either be connected to Penn's on-campus network or logged into the GlobalProtect VPN if accessing remotely. Once connected, PennChat can be accessed at https://pennchat.upenn.edu/
Q: Why do I get a blank screen when I try to access PennChat from off-campus?
A: To use PennChat from off-campus, you must first log in to Penn’s GlobalProtect VPN. Connect to the VPN at https://vpn.upenn.edu/ Once you are successfully connected, PennChat can be accessed at https://pennchat.upenn.edu/
Q: Which AI models are available through PennChat?
A: PennChat currently provides access to Large Language Models (LLMs) from OpenAI and Anthropic, both hosted within a Penn-provided network to ensure security and data privacy. PennChat provides secure access to Claude models in advance of a Penn Enterprise Agreement by leveraging the protections of Amazon Web Services’ Bedrock platform.
Q: Is my data safe when using PennChat?
A: Yes. Unlike many commercial AI tools, any data you enter into PennChat will not be used to train AI models. The service is hosted within Penn's secure network environment on AWS to help protect your information. Penn’s Data Risk Classification Policy is located at: https://isc.upenn.edu/security/penn-data-risk-classification
Q: What types of data can I use with PennChat?
A: You should always check with your local school or center’s IT Support Provider before entering data into any third-party tool. PennChat has been cleared for use with Low, Moderate, and most High Risk data. However, you should never enter Social Security Numbers (SSNs) or credit card numbers into the system, regardless of the context. Please avoid entering identifiable information (including names and Medical Record Numbers) when not specifically needed to complete the task.
Q: How can I select which of the AI models that I want to use?
A: Your default LLM model will be Claude Sonnet 4.6 from Anthropic. To change models, go to the top left corner of your main portal screen and click on the current model name (Sonnet 4.6) and select a different model from the drop down. Here is a popular LinkedIn Learning training course on chat prompting.
Q: Why isn’t the information I’m getting back the most up to date?
A: All LLMs are trained on data collected up to a certain point. When they’re released with a version number, 3.5 or 4.6, each of those versions has a specific cutoff point. For privacy reasons, your query is not sent out to the internet to look at the most recent information available. You can actually ask the PennChat model you’re using what its cutoff date is.
Q: How do I manage my balance on PennChat?
A: Every user is allocated a certain number of daily credit tokens to use to interact with different LLM models. Here are some tips for managing your balance:
- When choosing which LLM model to use, be aware that the ‘economical’ models use fewer tokens, and the ‘premium’ models use more.
- You can check your balance by clicking on your user avatar in the lower right hand corner of the PennChat screen.
- You can see more balance details in PennChat Settings. Click on your avatar in the lower left corner, choose settings, then click on the balance option. There you can see your current balance plus when the next refresh will occur.
We do not currently support a permanent increase in individual balances.