Artificial intelligence news.
Straight from the original announcements.

OpenAI reports healthcare results for computer-use agents

November 16, 2026, 07:00By the AI News Bites Editorial TeamSource: OpenAI

OpenAI is sharing new results on computer-use agents for healthcare, including clinical safety, helpfulness and administrative workflow performance. The company is also detailing work with healthcare organizations using agents to navigate existing systems for authorizations, referrals, discharge planning and member outreach.

Key points
  • OpenAI reports 99% safety ratings across 325 reviewed clinical tasks.
  • GPT‑6 Astra earned 94.3% of available OnHealthAdminBench points.
  • GPT‑6 Luna cost 98% less at 8.5 cents per task.
  • CVS Health, Abridge, Cigna Healthcare and Ambience are developing healthcare applications.
  • Organizations can configure approvals, access controls and audit logs.

Agents designed for fragmented healthcare workflows

Healthcare teams often move information manually between electronic health records, insurer portals and other applications. OpenAI’s computer-use agents are designed to carry out multi-step workflows by clicking, typing and moving between systems, including software that does not fully support APIs.

In a demonstration using synthetic patient data, an agent reviews a patient’s visit notes, physical therapy records and prior imaging after the patient reports lower back pain. It drafts an MRI order, an encounter note and a patient reply for the clinician to review, then prepares a prior authorization request in the insurer’s portal for an authorization specialist.

Healthcare teams, developers and individual clinicians can access these capabilities through the OpenAI API, ChatGPT for Healthcare and ChatGPT for Clinicians. Organizations can configure approvals and agent access, record audit logs and support HIPAA-compliant workflows.

OpenAI reports safety and performance results

OpenAI evaluated GPT‑6 Astra on tasks clinicians handle in their EHR inboxes. The tasks were shaped by physician input and tested clinical reasoning over longitudinal patient context as well as navigation of complex interfaces.

Reviewers rated 99% of 325 reviewed tasks as safe. Physicians rated the work “very helpful” or “extremely helpful” in 93% of 646 assessments. OpenAI says reviewers considered whether inaccuracies, omissions or actions could plausibly lead to patient harm.

For administrative workflows, including prior authorizations, appeals and medical equipment orders, GPT‑6 Astra earned 94.3% of available rubric points on OnHealthAdminBench, an independent benchmark from Stanford researchers. GPT‑6 Luna earned 91.8%, while costing 98% less at 8.5 cents per task. Both models were evaluated at maximum reasoning effort.

Healthcare organizations are testing computer use

CVS Health is using OpenAI’s capabilities for tasks including sending documents, processing referrals and supporting specialist consults. OpenAI says one production workflow cleared a months-long document-sending backlog over a weekend, while another handles the system steps for tens of thousands of consults each day. CVS Health uses validation checks, human oversight, screenshots and action records, and is open sourcing its Agent Kit, built with OpenAI’s Agents SDK.

Abridge is developing agents to coordinate care plans across applications, including discharge barriers, medication access and home-health coordination. Cigna Healthcare is using computer use to gather information from more than 10 separate clinical and administrative systems, giving case managers more context for member outreach. Ambience has built controls that pause an agent when a record is ambiguous and require clinician approval before submission.

Controls remain central to deployment

OpenAI says healthcare deployments require clear limits on agent actions, protection for patient information and human review. ChatGPT for Healthcare provides role-based access controls and audit logs, while organizations can add approval steps and restrict the applications, environments, records and actions available to an agent.

In the source’s examples, clinicians and specialists remain responsible for decisions requiring judgment, and agents can be paused or stopped. OpenAI also says organization data is not used to train its foundation models. Organizations seeking HIPAA-compliant workflows involving protected health information must use eligible services covered by their Business Associate Agreement.

Related news

Snorkel AI says gated training lifts Qwen3.8-27B on professional tasks
ResearchNovember 19, 2026, 15:32

Snorkel AI says gated training lifts Qwen3.8-27B on professional tasks

Snorkel AI says it raised Qwen3.8-27B's pass@1 performance on OpenAI's GDPval gold set from 31.9% to 40.7% by combining expert-built training data with binary-gated…

Snorkel AI boosts Qwen3.8-27B on difficult professional tasks
ResearchNovember 19, 2026, 15:32

Snorkel AI boosts Qwen3.8-27B on difficult professional tasks

Snorkel AI says it raised Qwen3.8-27B's pass@1 score on OpenAI's GDPval benchmark from 31.9% to 40.7% by combining expert-built training data with a modified…

Products and tools
Products and toolsNovember 16, 2026, 07:00

OpenAI adds computer-use agents and plugins to ChatGPT for Healthcare

OpenAI is expanding ChatGPT for Healthcare with computer-use agents, improved clinical search and more than 20 healthcare plugins. The company says the updates are…