US Public Health Agencies to Test OpenAI and Anthropic AI Tools
Public health agencies across the United States will begin testing generative artificial intelligence tools from OpenAI and Anthropic through a new programme designed to examine how AI can support government healthcare operations.
The initiative, called the Public Health Use Case and Learning Scaling Engine, or PULSE, is being led by the Coalition for Health AI in partnership with OpenAI, Anthropic, and Accenture.
Ten state, local, tribal, or territorial public health jurisdictions will participate in the programme. Their experiences will be used to develop practical implementation playbooks for other government health agencies considering similar AI deployments.
OpenAI and Anthropic have donated ten enterprise licences capable of supporting as many as 2,000 public health professionals. Accenture will handle participant onboarding and assist in documenting lessons from the trials.
However, the coalition has not yet disclosed which OpenAI and Anthropic products, model versions, or configurations will be deployed. It also remains unclear how the two AI providers will be assigned across the participating agencies.
“Every major technological transformation succeeds or fails based on trust, governance and execution,” said Dr. David Lakey, a former Texas health commissioner.
Lakey said PULSE was created to focus on practical implementation rather than simply demonstrating what AI can do.
Five Public Health Applications
Participants will be grouped into communities focused on five proposed AI use cases:
- Biosurveillance and drug-wave prediction
- Social determinants of health mapping
- Government operations, efficiency, and community-feedback analysis
- Public communications and multilingual translation
- Automated clinical-data retrieval using a FHIR query engine
FHIR, or Fast Healthcare Interoperability Resources, is a standard developed by HL7 for electronically exchanging healthcare information between compatible systems.
The inclusion of a FHIR query engine suggests that AI could help practitioners locate, retrieve, or summarise clinical information. However, the announcement does not clearly define whether the models will generate database queries, retrieve records, interpret results, or perform all three functions.
There is also no published explanation of how incorrect queries, incomplete records, unsupported summaries, or misleading AI-generated conclusions will be detected before the information is used.
Questions About Privacy and Patient Data
Several of the proposed applications could involve clinical, demographic, geographic, or population-health information.
The Coalition for Health AI has not stated whether the pilot projects will use identifiable patient records, anonymised information, aggregated datasets, or entirely synthetic data.
This distinction is important because some participating organisations and workflows may fall under the US Health Insurance Portability and Accountability Act, or HIPAA.
Under HIPAA, covered healthcare organisations and their service providers must protect electronic health information when it is created, received, stored, or transmitted through cloud-based systems.
However, HIPAA may not apply to every agency or activity involved in PULSE. Its application will depend on the type of organisation, the information being processed, and how the AI system is used.
OpenAI says inputs and outputs from its business services, including ChatGPT Enterprise and its API, are not used to train its models by default.
Anthropic has issued a similar policy for its commercial products.
Still, these general provider policies do not explain how the PULSE deployments will be configured. Details regarding data retention, storage locations, access controls, audit logs, cybersecurity protections, and the handling of protected health information have not yet been made public.
“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 said the enterprise licences would give public health organisations an opportunity to evaluate AI through a structured process.
Human Oversight Still Undefined
Another major unanswered question is how AI-generated outputs will be reviewed by trained personnel.
The announcement does not specify whether public health staff will be required to:
- Approve AI-generated public advisories before publication
- Verify multilingual translations
- Validate clinical information retrieved by AI
- Review biosurveillance predictions
- Confirm drug-wave forecasts before operational use
The US National Institute of Standards and Technology recommends evaluating AI systems according to their intended purpose, operating environment, affected communities, and potential consequences.
NIST guidance also recommends identifying which AI functions require human oversight and ensuring that users understand the limitations and performance of the systems they operate.
For generative AI, this includes testing, monitoring, documentation, privacy safeguards, validation procedures, and management accountability.
“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.
Kelly said PULSE would allow public health professionals to test AI tools inside their own working environments while incorporating privacy, governance, and responsible-use measures.
AI Adoption Remains Limited
Data from the National Association of County and City Health Officials, cited by the Coalition for Health AI, indicates that nearly 40 percent of local health departments were not yet using AI.
Some agencies have expressed interest in using AI to modernise workflows, analyse public feedback, improve communications, and reduce administrative pressure on understaffed teams.
PULSE will provide participating agencies with enterprise AI access, onboarding assistance, peer-learning communities, and implementation guidance.
However, minimum requirements involving staffing, technical infrastructure, cybersecurity, data interoperability, and internal AI expertise have not yet been released.
Eligible participants include:
- State and territorial health departments
- County and municipal health agencies
- Tribal public health authorities
- Indian health organisations
- Health departments serving large cities
The coalition plans to convert findings from the ten participating jurisdictions into guidance that can be adopted by other public health agencies.
Still, public health systems differ widely in size, funding, technical capacity, legal authority, staffing, and procurement procedures. PULSE has not yet explained how its final playbooks will account for these differences.
Pilot Testing Begins in Late 2026
The first PULSE trials are expected to begin in autumn 2026. Implementation playbooks based on the programme are scheduled for release in 2027.
The Coalition for Health AI has not yet published the performance indicators that will be used to evaluate the pilots.
It is also unclear whether each use case will be assessed using separate standards for technical accuracy, privacy, safety, accessibility, operational value, and human oversight.
The programme forms part of the coalition’s broader efforts to establish governance standards for artificial intelligence in healthcare.
In May, the organisation announced plans to develop guidance covering eight areas of AI governance through workshops and working groups involving more than 150 representatives from the healthcare and technology sectors.
It has since begun releasing playbooks covering organisational AI policies, governance structures, and internal resources.
The coalition has also worked with the Joint Commission on governance playbooks connected to its voluntary Responsible Use of AI in Healthcare certification. The PULSE announcement, however, does not indicate whether participating public health agencies will be assessed under that certification.
Dr. Brian Anderson, chief executive of the Coalition for Health AI, said many public health organisations entered the COVID-19 pandemic after years of limited investment in digital technology.
He said PULSE would give public health professionals direct experience with AI before governments consider deploying the technology more widely.
“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, a former White House COVID-19 response coordinator.
Jha said the programme would help determine which public health applications are effective and document the findings for other agencies.
Why This Matters to the Philippines
The PULSE programme could provide useful lessons for Philippine institutions such as the Department of Health, local government health offices, public hospitals, disease-surveillance units, and emergency-response agencies.
AI could potentially assist Filipino public health workers in analysing disease trends, translating health advisories into Filipino and regional languages, reviewing community feedback, and retrieving information from fragmented health databases.
However, the same concerns raised in the United States also apply to the Philippines.
Before AI is used in public healthcare, agencies must clearly define how patient information will be protected, who will be accountable for incorrect outputs, how generated translations will be verified, and when a licensed health professional must review the system’s recommendations.
AI can help public health teams work faster, but it cannot replace professional judgment, strong data-governance policies, secure digital infrastructure, and human accountability.
The real test of PULSE will not be whether OpenAI and Anthropic models can produce impressive answers. It will be whether public health agencies can use them safely, accurately, transparently, and responsibly in real-world government operations.
