Anthropic Enterprise FAQ

Penn’s Anthropic Enterprise offers AI tools tailored for large-scale operations with Penn-specific protections included. The offering includes Claude Chat and Claude Code and is approved for Low and Moderate risk data.  This offering is not approved for HIPAA covered entities, but a high risk data version will be available at a later date.

Anthropic’s LLMs do not train on Penn data, among other protections.  

To request this service, please ask your local IT organization to open a ticket with ISC. They will need your PennKey and 26-digit budget code. You also need to request whether you would like Standard ($20/month) or Premium ($50/month).  

Purchasing a license requires a commitment for the fiscal year.  Your monthly charges begin in the month you request the license.  If you start with a standard license, you can upgrade during the fiscal year.   If you have a premium license, you will not be allowed to downgrade during the fiscal year.  

We are offering two service levels: Standard ($20/month) and Premium ($50/month).  Standard includes up to 112 Claude messages per user within any rolling five-hour period, and a $50 weekly API usage allocation at standard API pricing. Premium increases those limits to 225 messages per user within any rolling five-hour period, and a $250 weekly API usage allocation.  

Beyond usage caps, Standard and Premium are identical with one exception: a Premium license is currently required to access the Fable 5 model.

Standard is the right starting point for most people. Premium makes sense if you expect to use Claude heavily on a given day — long documents, large coding sessions, or extended back-and-forth work — since it roughly doubles both the message and dollar allowance. If you have never used Claude, elect the Standard plan.  

Purchasing a license requires a fiscal year commitment.  If you start with a standard license, you can upgrade during the fiscal year.   If you have a premium license, you will not be allowed to downgrade during the fiscal year.

Once you hit your usage limit, you'll see a notification saying you've reached it, and you won't be able to send new messages until it resets on a rolling 5-hour cycle. Your existing conversation stays visible (you can still read/copy from it), but no further requests get processed.

To see more detail, go to your Account menu → Settings → Usage. There, you can see how much of your plan's five-hour session limit you've used so far, along with the time remaining in the session.

We are not currently offering the opportunity to purchase extra credits.  We will evaluate adding this option in the future based on demand.  You can upgrade from a standard to a premium license for a higher allocation, or use a less expensive model to conserve your credit allocation.

Claude Code is available. Additional Claude AI tools (Science, Design, etc.) are not currently available as part of our Claude Enterprise license. We are exploring how we can include them in the future.

Claude CoWork is currently not enabled, but it is being reviewed for privacy controls, and we hope to be included in the offering before the end of the year.

Penn's current offering covers Low and Moderate risk data under Penn's data risk classification. A HIPAA-compliant, high-risk-data version is planned but isn't available yet. For now, please do not put PII, PHI or other high-risk data into Claude. We will announce when a High Risk-certified version is available.  It is important to be aware of the risks associated with different types of data. If you are a student, contact your course instructor for guidance. If you are a faculty member, contact your department chair or dean. Please refer to Penn’s Office of Information Security’s Statement on Guidance for the University of Pennsylvania Community on Use of Generative Artificial Intelligence for additional guidance.  

This offering is not approved for HIPAA covered entities, but a high risk data version will be available at a later date.  When that offering is available, this FAQ will be revised accordingly.

The best model for you depends on your needs.  We suggest following Anthropic’s guide for choosing the right model for you.The best model for you depends on your needs.

• Haiku is fastest and cheapest for quick, simple requests
• Sonnet is the strongest all-around choice for most day-to-day work
• Opus is more capable but more expensive per message
• Fable (Premium Tier Only) is a top-level model built for the most ambitious, long-running projects.  It is the most expensive model to run.

 We suggest following Anthropic’s guide for choosing the right model for you.

Use one of the following options:

• Web
• Apps are available for iOS, Windows, Android, and Mac OS.  Be aware that Penn-managed devices may have restrictions on what can be installed.

Log into Anthropic using by entering your email, and choosing the SSO option to sign in.  Then, log in with your PennKey.  If you need assistance installing one of the Anthropic apps, please contact your local IT organization

Penn Medicine, including the Perelman School of Medicine, operates in a HIPAA-covered environment and maintains specific requirements for technology platforms that may be used within its workforce. While Penn's Anthropic Claude enterprise service has been approved for certain University use cases, it has not been approved for use within Penn Medicine at this time. Penn Medicine continues to evaluate AI technologies and will share additional guidance as new tools and capabilities are reviewed and approved.

Penn Medicine faculty, staff, students, and trainees should use only AI tools that have been reviewed and approved for use within the Penn Medicine environment. Approved tools, available use cases, and implementation guidance are maintained on the Penn Medicine AI Hub. Microsoft Copilot Chat continues to be available for use by all Penn Medicine (including PSOM and UPHS) faculty, staff, and students via their Microsoft 365 account.