OpenAI Introduces GPT-6 Computer-Use Agents for Healthcare Administration
OpenAI has introduced GPT-6 Astra and GPT-6 Luna to handle healthcare administrative tasks, navigating clinic software and insurance portals like human clerks.

OpenAI is targeting the widespread lack of standardized APIs in the healthcare sector by deploying computer-use agents designed to navigate existing clinical software and insurance portals. The new models, designated as GPT-6 Astra and GPT-6 Luna, operate user interfaces in a manner similar to human staff to bridge gaps between electronic health records and external web portals.
According to OpenAI, these agents can assist clinicians when a patient reports symptoms like back pain by searching previous findings, drafting MRI orders, composing patient replies, and preparing prior authorization requests for insurers. All drafts require review by clinicians or administrative staff before final release. The tools are accessible via the Agents API as well as dedicated platforms like ChatGPT for Healthcare and ChatGPT for Clinicians.
Performance and Real-World Deployments
In evaluations detailed by OpenAI, GPT-6 Astra processed electronic health record inbox tasks—such as determining ibuprofen dosages for a child with a broken arm—with physician reviewers rating 99 percent of 325 evaluated tasks as clinically safe. Reviewers also judged the preparation work as very or extremely helpful in 93 percent of 646 reviews.
For administrative workflows like prior authorizations, appeals, and medical equipment orders, OpenAI referenced the independent HealthAdminBench benchmark from Stanford University researchers. GPT-6 Astra achieved 94.3 percent of possible points, while the budget-focused GPT-6 Luna scored 91.8 percent at a cost of approximately 8.5 cents per task, representing a 98 percent price reduction compared to Astra.
Several major US healthcare entities are already testing and deploying these systems. At CVS Health, agents manage referrals and specialist consultations across disjointed systems, clearing a months-long backlog of outgoing documents over a single weekend and handling steps for tens of thousands of daily consultations. CVS has also published its software architecture on GitHub under the name Agent Kit. Similarly, Cigna Healthcare utilizes the models to unify patient details from over ten disparate systems, while startups like Assort integrate the features to execute final clicks in practice management software lacking official API support.
Privacy, Security, and Compliance
Because these digital agents interact directly with confidential health records, data protection remains a central focus. OpenAI states that it does not use organizational data to train its base models. ChatGPT for Healthcare incorporates role-based access controls and audit logs, allowing organizations to restrict agent access strictly to necessary records and actions.
For workflows subject to HIPAA regulations regarding protected health information, organizations must execute a Business Associate Agreement (BAA) with OpenAI, utilize covered services, and maintain mandated security configurations. US medical professionals utilizing the individual ChatGPT for Clinicians plan can also sign a BAA covering specific features. CVS Health maintains data within its own controlled infrastructure, overseeing agent actions via screenshots and logs, though OpenAI's announcements did not specify processing locations for screenshots or address compliance with Europe's GDPR, keeping the current rollout focused on the US market.
What it means for developers
Developers working on enterprise software integration can leverage these developments to bridge gaps where traditional APIs are absent. By utilizing computer-use agents, applications can interact reliably with legacy desktop or web interfaces through visual actions rather than direct backend integration. For those looking to build and test similar capabilities, developers can try top AI models cheaply through one API at https://apixoai.online. As this technology scales, ensuring robust monitoring, clear human oversight for critical decisions, and strict data governance will remain vital considerations for engineering teams building autonomous workflows in regulated industries.
Source: AI agents are starting to fill out medical forms and file insurance requests in place of human staff — The Decoder. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key

