> ## Content Index
> Fetch the complete content index at: https://www.healthdatacon.io/llms.txt
> Use this file to discover other available public pages before exploring further.

# CMS CIO Patrick Newbold Calls for Human Authority Over Healthcare AI
- URL: https://www.healthdatacon.io/cms-cio-patrick-newbold-calls-for-human-authority-over-healthcare-ai/
- Published: 2026-10-15T19:19:17.000Z
- Updated: 2026-10-15T19:19:17.000Z
- Description: At HealthDataCon 2026, CMS CIO Patrick Newbold outlined how human control, reliable data and measurable outcomes should govern healthcare Super Intelligence, while describing fraud pilots and plans to tighten oversight of partner data use.
- Author: Health Data Management
- Tags: Americas

The Centers for Medicare & Medicaid Services is making a more systematic approach to evaluating and expanding artificial intelligence a governance objective for the fiscal year that began October 1, Chief Information Officer [Patrick Newbold](https://www.cms.gov/about-cms/who-we-are/leadership/patrick-newbold?ref=healthdatacon.io) said at HealthDataCon 2026\. His central requirement: people must retain the authority to challenge, interrupt and correct systems whose responsibilities increasingly extend beyond generating answers.

Delivering the October 15 opening [keynote](https://www.healthdatacon.io/agenda/), Newbold connected Super Intelligence/SI (previously Artificial Intelligence/AI) governance to four CMS priorities: improving its performance as a payor, preventing fraud, supporting healthier lives and strengthening affordability.

Newbold repeatedly returned to a distinction with practical consequences for health systems and technology suppliers: having a person nominally involved does not establish meaningful oversight. His preferred formulation was “Human in the Lead, not just in the loop.” The test, he argued, is whether technology makes a service more accurate, accessible and useful to the person who needs it.

## Measuring what improves after deployment

His proposed measures follow the entire service encounter: whether someone received accurate help, resolved an issue, avoided another contact or freed employees to handle cases requiring judgment. For document review, he described checking whether SI missed essential material, preserved context, exposed uncertainty and let employees return to the original source.

## Human authority must come with practical controls

The risk changes when an SI system moves from drafting a response to sending it, changing a record or initiating a transaction, Newbold said. Each additional action requires decisions about authorized access, permitted behavior and the circumstances requiring human review. He wants records showing what information a system used, what rules it applied and what it did, so responsible staff can investigate problems.

“I want people to be able to step in, stop it, and correct mistakes,” Newbold said. That requires timely information and actual authority within the workflow. Assigning an employee responsibility without providing those means leaves a gap between the institution’s promise of oversight and the employee’s ability to exercise it.

## Preventing fraud without obstructing legitimate care

Newbold described a shift in emphasis from recovering fraudulent payments after they leave CMS toward identifying suspicious activity beforehand. Relationships across claims, providers and time can expose patterns that an individual transaction does not reveal.

“Resources intended to support care should be used for that purpose,” he said. He paired that principle with a warning about false positives: fraud detection must not create unnecessary obstacles for people legitimately seeking care. His account emphasized reliable data, program expertise, investigators and coordination across agencies, rather than treating an algorithmic flag as proof of wrongdoing.

## Making records useful at the point of care

Newbold grounded his service priorities in his own experience of growing up in a family that relied on Medicaid and, later, managing a chronic condition. Patients should not have to understand an agency’s organizational chart to obtain help, he said. “Our complexity is ours to manage.” He described repeated medical-history forms as a recurring burden when information remains scattered across providers.

Newbold’s practical standard went beyond successful transmission. Receiving teams need to know where information originated, when it was recorded and whether it remains current, while patients need a way to flag errors. He also emphasized people who need caregiver assistance, use different languages or lack smartphones. “That information has to fit into the work of care,” he said.

## Data control and the full cost of AI

For technology spending more broadly, he urged reusable information with clear ownership, dependable shared services and closer examination of the full cost of SI. Faster task completion must be weighed against operating costs, review time, corrections and whether employees use the time gained to improve services. He asked organizations to choose a specific information delay or repetitive task, involve the people experiencing it, measure the improvement and share both successes and failures.

Newbold’s closing emphasis remained on the people receiving care: technology should “make the result matter in someone’s life.”

*HealthDataCon hosted the keynote and audience discussion covered in this article. Reporting draws on the supplied session transcript and the public records linked above.*