PULSE Gives 10 US Health Agencies OpenAI And Anthropic AI Pilots

Ten US public health jurisdictions will test enterprise AI tools from OpenAI and Anthropic under the PULSE programme, AI News reported, putting donated licences, public-health workflows and unresolved governance rules inside the same pilot.

The Public Health Use Case and Learning Scaling Engine is being run with the Coalition for Health AI, OpenAI, Anthropic and Accenture.

The donation covers 10 enterprise licences with capacity for up to 2,000 public health practitioners, while Accenture will handle participant onboarding and help turn the trials into implementation playbooks.

PULSE Gives Agencies Five AI Workflows To Test

The Coalition for Health AI leadership council will choose the participating state, local, tribal or territorial jurisdictions and assign practitioners to communities built around five use cases.

AI News identified the workstreams as biosurveillance and drug-wave prediction; social determinants of health mapping; operations and community-feedback review; public communications with multilingual translation; and clinical-data retrieval through a FHIR query engine.

The five-use-case design takes the pilots into data-heavy public health work rather than a single chatbot test.

FHIR is the HL7 standard used to exchange healthcare information electronically between compatible systems, and one PULSE use case involves clinical-data retrieval through that format.

The public record does not define whether generative AI will generate queries, retrieve records, summarise returned information or combine those functions.

The US Department of Health and Human Services requires HIPAA-covered organisations to safeguard electronic protected health information.

HIPAA coverage will vary by agency, dataset and workflow, so the pilots cannot be treated as one uniform compliance case.

Governance Rules Remain Separate From The Pilot Scale

Dr David Lakey, a former Texas health commissioner, framed trust, governance and execution as the conditions that determine whether major technology shifts succeed.

Coalition for Health AI chief executive Dr Brian Anderson associated the programme with public-health technology underinvestment before COVID-19 and presented the pilots as a way for agencies to gain practical AI experience before broader implementation.

The PULSE timetable schedules the programme to begin in autumn 2026.

CHAI expects the 10-jurisdiction trial to produce playbooks in 2027 for other public health agencies considering similar deployments.

NIST's AI Risk Management Framework recommends evaluating AI systems by intended use, operating environment, affected parties and potential consequences.

CHAI has not published use-case-specific requirements for evaluation, privacy, security or human review, a material gap because some applications could involve demographic, geographic, clinical or population-health information.

The trial design also leaves procurement and operating readiness unresolved.

The eligibility pool covers state and territorial departments, city and county agencies, tribal authorities, Indian health organisations and large-city public health bodies, but the public record does not set a minimum technology stack or staffing baseline for those organisations.

CHAI's May governance effort covers eight governance areas and uses workshops and working groups with more than 150 healthcare AI representatives.

The PULSE announcement does not say participating public health agencies will be assessed under the Joint Commission-aligned voluntary certification work.

OpenAI And Anthropic Policies Do Not Define The Pilots

OpenAI's business-service policy excludes ChatGPT Enterprise and API content from model improvement by default.

Anthropic gives a similar default training boundary for commercial-product inputs and outputs.

Those vendor policies do not specify how the PULSE deployments will be configured, retained, audited or limited for public health data.

Elizabeth Kelly, Anthropic's head of beneficial deployments, said public health teams are being asked to do more with less and that AI can help if introduced with care and guardrails.

Felipe Millon, OpenAI's head of government go-to-market, presented the donated licences as a way for public health organisations to evaluate the tools through a structured process.

The pilot scale contrasts with unresolved deployment rules.

CHAI cited National Association of County and City Health Officials data showing that nearly 40% of local health departments were not using AI, while the PULSE design gives selected agencies enterprise licences, onboarding, peer communities and implementation playbooks.

The same record does not specify minimum staffing, infrastructure, interoperability or cybersecurity requirements for participating jurisdictions.

The PULSE announcement leaves the model versions, provider assignment method and human-review rules outside the public record.