PennChat: Secure AI Portal with Claude, ChatGPT Models at Penn

Penn has started piloting an artificial intelligence portal called PennChat, offering secure access to multiple large language models. The service rolled out to the university community on July 7th. Here's the thing: PennChat includes specific Claude and ChatGPT models. It's hosted within Penn’s secure network environment on Amazon Web Services. This setup means companies cannot use Penn data to train AI models. Information Systems and Computing emphasizes that PennChat is "cleared for use with Low, Moderate, and most High Risk data." However, users should not enter Social Security numbers, credit card numbers, or protected health information. What's interesting is that the interface’s large language models cannot send queries to the internet. They are only reliably trained on data collected before cutoff dates ranging from 2024 to 2025. Users need to be connected to PennNet, AirPennNet, or the University’s GlobalProtect VPN to access the portal. They can select from 14 different models, categorized by credit consumption. Each user receives a daily refreshable balance of credits, with some users currently allotted 1.85 million credits. The platform supports file uploads and can generate Word, Excel, CSV, and PDF documents. Only a small group of administrators can view chat history for troubleshooting purposes. The bottom line: This initiative aims to provide safe and equitable AI tool access, which could revolutionize teaching, research, and business at Penn.