WashU Generative AI Environment Update

WashU Digital Transformation and WashU IT have been hard at work further building out the WashU Generative AI Environment and are pleased to announce two new enhancements.  

WashU ChatGPT Beta upgrade 

Recently, the Washington University ChatGPT Beta was upgraded to use Open AI’s new GPT-4o “omni” model – one of their most advanced artificial intelligence model capable of complex problem solving.  

The upgrade to GPT-4o will provide better functionality to WashU users, with text generation at double the speed and significant improvements with non-English languages. Additionally, it will offer a more seamless user experience when moving between the secure sandbox and the current public version. 

Secure API Beta launch 

A limited number of API endpoints within WashU’s secure generative AI environment are now available as part of a beta release designed to quickly make this higher-level GPT functionality accessible for use with sensitive data.  

Just like web browsers allow you to access different websites by entering URLs, API endpoints enable different software applications or services to communicate and exchange information. They are used by software programs to retrieve or send data to specific parts of another program or service, enabling applications to connect and collaborate, and exchange or request information. 

Endpoint services will be billed at cost based on usage. A custom-built feature allows users to set a fixed budget for API endpoint usage and avoid surprisingly high bills.  

Faculty, staff and students who wish to use API endpoints in pursuit of the university’s mission may now find user documentation and request access.   

Other generative AI tools 

Additional tools outside of our own secure sandbox, like Zoom’s AI Companion, have also been approved as fully secure for use with sensitive data, including information covered by HIPAA and FERPA. A current list of AI-enabled tools and whether they are approved for use is available on the IT website. This list identifies if the tool is appropriate for use with sensitive data and to what extent.  

It is critical to never enter sensitive information in a nonapproved tool. 

In line with the university’s procurement policy for IT hardware and software, vendor generative AI tools must be assessed for risk by WashU IT’s Office of Information Security prior to use. Please visit our procurement page for more information. Assessment is required for both existing tools that add or expand AI capabilities and new purchases of vendor AI tools. Complete the AI Interest Form to submit a request for review. You will be prompted to enter your email address before you can access the form. 

We encourage you to learn more about generative AI tools at our new AI hub website, and ask questions in the Research Town Hall this afternoon.