US health agencies evaluate OpenAI and Anthropic AI models

Public health agencies nationwide are set to evaluate generative AI through a new initiative led by the Coalition for Health AI, in partnership with OpenAI, Anthropic, and Accenture.

The Public Health Use Case and Learning Scaling Engine (PULSE) will fund trials across 10 state, local, tribal, and territorial jurisdictions. The initiative aims to generate actionable implementation guidance for public health organizations considering similar AI deployments.

OpenAI and Anthropic have contributed 10 enterprise licenses, supporting up to 2,000 public health professionals. Accenture will manage participant onboarding and assist in developing implementation playbooks based on trial results.

The program grants public health staff access to enterprise-grade AI products from OpenAI and Anthropic. However, CHAI has not disclosed the specific products, model versions, or configurations to be used, nor how the providers will be allocated across the pilot sites.

“Every major technological transformation succeeds or fails based on trust, governance and execution,” said Dr. David Lakey, former Texas health commissioner. “PULSE is designed to support agencies in this effort, focusing on practical implementation.”

Five public health use cases

The Coalition for Health AI’s leadership council will select participating jurisdictions and assign practitioners to five key areas: biosurveillance and drug-wave prediction, social determinants of health (SDoH) mapping, operational efficiency and community feedback analysis, public communications, and a multilingual translation hub. Additionally, the program includes automated clinical data retrieval and a FHIR query engine.

The NIST AI Risk Management Framework advises evaluating AI systems based on their intended use, operating environment, affected parties, and potential consequences. CHAI has not yet published specific evaluation, privacy, security, or human-review criteria for these five PULSE use cases.

FHIR is an HL7 standard facilitating the electronic exchange of healthcare information between compatible systems.

While PULSE includes an automated clinical data retrieval and FHIR query engine, the announcement does not clarify the role of generative AI in this workflow. It remains unclear whether the models will generate queries, retrieve records, summarize information, or combine these functions.

Furthermore, the announcement does not specify whether incorrect queries, incomplete retrievals, or unsupported summaries will be reviewed by staff before the information is utilized.

Some proposed applications may involve demographic, geographic, clinical, or population health data. CHAI has not indicated whether the pilots will use identifiable records, de-identified information, synthetic data, or aggregated datasets.

Under HIPAA, organizations covered by the regulation must protect certain electronic health information. The Department of Health and Human Services’ cloud-computing guidance mandates that regulated entities and service providers comply with HIPAA requirements when cloud systems create, receive, maintain, or transmit electronic protected health information.

HIPAA regulations will not apply to all participating organizations or workflows. Their applicability depends on the specific agency, the data involved, and the function being performed.

OpenAI states that inputs and outputs from its business services, including ChatGPT Enterprise and its API, are not used to train or improve its models by default. Similarly, Anthropic confirms that it does not use inputs and outputs from its commercial products for model training by default.

These provider policies do not define the configuration of PULSE deployments. The announcement lacks details on retention periods, access controls, audit arrangements, data-storage requirements, or rules for submitting protected health information.

“We believe AI should be useful, safe and accessible to the people tackling society’s most important challenges,” said Felipe Millon, OpenAI’s head of government go-to-market. He noted that the licenses are intended to help public health organizations evaluate these tools through a structured process.

Governance and evaluation remain undefined

The pilots are scheduled to launch in autumn 2026. CHAI expects the resulting playbooks to be released in 2027, serving as reference material for other public health agencies.

CHAI has not published the metrics for assessing the pilots or explained whether each use case will be evaluated under separate technical, operational, privacy, and safety criteria.

The announcement also fails to explain how model outputs will be reviewed. It does not specify whether staff must approve generated public communications, verify translations, validate retrieved clinical information, or review biosurveillance and drug-wave outputs before use.

NIST guidance recommends identifying which AI functions require human oversight and training users to understand system performance and limitations. Its generative AI guidance also covers testing, validation, monitoring, documentation, privacy, and management oversight.

“Public health teams are being asked to do more with less, and AI can help — as long as it’s brought in with care and the right guardrails,” said Elizabeth Kelly, Anthropic’s head of beneficial deployments. She stated that PULSE allows practitioners to test tools in their own environments, incorporating privacy, governance, and responsible-use measures from the start.

Data from the National Association of County and City Health Officials, cited by CHAI, indicates that nearly 40% of local health departments are not currently using AI. The coalition noted that some departments are interested in revising workflows and improving operational efficiency.

PULSE will provide enterprise licenses, onboarding support, peer communities, and implementation playbooks. However, CHAI has not specified the minimum staffing, infrastructure, interoperability, or cybersecurity requirements for participating jurisdictions.

Eligible participants include state and territorial health departments, county and municipal agencies, tribal authorities, Indian health organizations, and large city health departments.

PULSE plans to convert findings from 10 jurisdictions into guidance for broader use. The announcement does not explain how the playbooks will account for differences in agency size, technical systems, legal responsibilities, staffing, or procurement arrangements.

Additionally, the announcement does not clarify whether outputs from biosurveillance, drug-wave prediction, or clinical data retrieval will be used solely for testing, presented to staff for review, or incorporated into operational workflows.

PULSE is part of the Coalition for Health AI’s broader work on governance standards for healthcare AI. In May, the organization announced plans to develop guidance covering eight governance areas through workshops and working groups involving over 150 healthcare AI representatives.

The coalition has begun publishing playbooks on organizational AI policies, governance structures, and internal resources. Additional guidance is expected to cover other areas of healthcare AI management.

Separately, CHAI has collaborated with the Joint Commission on governance playbooks aligned with its voluntary Responsible Use of AI in Healthcare certification. The PULSE announcement does not state that participating public health agencies will be assessed under this certification.

Dr. Brian Anderson, CEO of the Coalition for Health AI, noted that public health agencies entered the COVID-19 pandemic after years of limited technology investment. He stated that PULSE aims to provide agencies with practical AI experience before wider implementation.

“We know AI is going to reshape how we deliver public health — the question is whether we do it thoughtfully or not,” said Dr. Ashish Jha, former White House COVID-19 response coordinator. He added that the program will test which applications work and document findings for other agencies.

See also: Bunkerhill raises $55M to scale agentic AI across health systems

Want to learn more about AI and big data from industry leaders? Explore AI & Big Data Expo, taking place in Amsterdam, California, and London. This comprehensive event is part of TechEx and is co-located with other leading technology events, including the Cyber Security & Cloud Expo. Click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.