Optum AI chatbot was left publicly accessible

An internal AI chatbot developed by Optum for employee use was unintentionally left publicly accessible online, as reported by _TechCrunch_ on December 13. The SOP Chatbot, a demo tool active since September, was designed to assist employees in navigating Optum's standard operating procedures. Although it was not used in real-world applications and contained no protected health information, it could answer questions about handling patient health insurance claims and disputes based on Optum's internal documents. These documents were securely stored within UnitedHealthcare's network, requiring employee login for access, yet the chatbot could reference them in its responses.

The chatbot was hosted on an internal Optum domain, but its IP address was exposed, allowing unrestricted internet access without a password. An Optum spokesperson clarified to _Becker's_ that the chatbot was a proof-of-concept demo, never intended for production, and is no longer accessible. The demo aimed to test the tool's response to a small set of SOP documents, with no PHI involved in its training or use. The technology was not scaled or used in any real capacity, and the duration of its public accessibility remains unknown.

Key takeaways:

An internal AI chatbot developed by Optum was accidentally left publicly accessible online, allowing unrestricted access via a web browser.
The chatbot, called SOP Chatbot, was a demo tool designed to help Optum employees navigate the company's standard operating procedures.
Although the chatbot referenced internal documents about handling claims and reimbursements, it contained no protected health information and was not used in real-world scenarios.
The chatbot was never put into production, and its public accessibility was unintended; it has since been taken offline.

Optum AI chatbot was left publicly accessible

Key takeaways:

Comments (0)

Newsletter