What OpenAI and Anthropic’s Public Health Pilot Teaches Us About AI Governance.

PULSE Puts OpenAI and Anthropic Inside U.S. Public Health Agencies — But the Guardrails Are Still Being Written:

A 10-jurisdiction pilot will hand up to 2,000 public health practitioners enterprise AI access — while questions about data, oversight, and accountability remain open.

10: Jurisdictions in the PULSE pilot

2,000: Practitioners with enterprise AI access

40%: Local health departments not yet using AI

The Big Governance Gap inside U.S. Public Health’s New AI Engine:

Two of the biggest names in AI are about to get a real-world test inside the systems that track disease outbreaks, translate public health notices, and pull clinical records — and the rulebook for how that will actually work is still being drafted.

The Coalition for Health AI (CHAI), OpenAI, Anthropic, and Accenture have launched PULSE — the Public Health Use Case and Learning Scaling Engine — a pilot programme placing enterprise generative AI tools into 10 state, local, tribal, or territorial health agencies. OpenAI and Anthropic are donating 10 enterprise licences covering up to 2,000 practitioners, with Accenture handling onboarding and playbook development.

1: Why Public Health Is Turning to AI Now:

Data cited by CHAI from the National Association of County and City Health Officials found that nearly 40% of local health departments are not using AI at all. Coming out of a pandemic that exposed years of underinvestment in public health technology, CHAI frames PULSE as a chance for agencies to build practical experience with AI before wider rollout — rather than adopting it blind.

"Every major technological transformation succeeds or fails based on trust, governance and execution," said Dr. David Lakey, a former Texas health commissioner involved with the programme. Anthropic's head of beneficial deployments, Elizabeth Kelly, echoed that framing, saying the pilot would let practitioners test the tools "in their own environments with privacy, governance, and responsible-use measures incorporated from the start."

2: Five Use Cases, One Shared Question:

Participating practitioners will be grouped around five use cases: biosurveillance and drug-wave prediction, social determinants of health mapping, operations and community-feedback analysis, a multilingual public communications hub, and automated clinical-data retrieval through a FHIR query engine.

What isn't yet defined is how each of those workflows will actually be checked. The announcement doesn't say whether AI-drafted public communications, translations, retrieved clinical data, or biosurveillance outputs need to be reviewed by staff before they're used — a gap that matters most in the highest-stakes use case, automated clinical-data retrieval, where the announcement doesn't even specify whether the models will generate the queries, retrieve the records, summarise them, or all three.

"We know AI is going to reshape how we deliver public health — the question is whether we do it thoughtfully or not." — Dr. Ashish Jha, former White House COVID-19 Response Coordinator

3: The Governance Gap Nobody's Filled Yet:

CHAI hasn't published which model versions or product configurations will be used, how OpenAI and Anthropic will be assigned across the 10 jurisdictions, or what evaluation criteria each use case will be measured against. NIST's AI Risk Management Framework calls for evaluating systems by their intended use, environment, affected parties, and consequences — but CHAI has not yet released separate privacy, security, or human-review requirements for any of PULSE's five use cases.

HIPAA compliance is another open thread. Both OpenAI and Anthropic say business-tier inputs and outputs aren't used to train models by default, but that policy doesn't answer how each PULSE deployment will handle retention periods, access controls, audit trails, or the submission of protected health information — and HIPAA itself won't apply uniformly across every participating agency and workflow.

Pilots are set to begin in autumn 2026, with CHAI expecting to publish implementation playbooks in 2027 for other public health agencies to reference.

4: What This Signals for Every Organisation Weighing AI Adoption:

PULSE is a useful case study well beyond public health. It shows two of the sector's biggest platforms being deployed into high-stakes, regulated environments before the governance framework around them is finished — oversight rules, data-handling standards, and success metrics are all being built in parallel with the rollout rather than ahead of it.

For most organisations, that order of operations is backwards — and avoidable. The gaps CHAI is still working through — who reviews AI outputs, how data is handled, what "successful" actually looks like — are exactly the questions that should be answered before an AI tool goes live, not after.

Deploy AI With the Guardrails Built In From Day One.

PULSE proves that even the biggest names in AI are still figuring out governance in real time. Otherworlds AI's Agent+ platform gives your organisation a structured, accountable AI deployment from the start — clear oversight, defined workflows, and enterprise-grade support, without the wait for someone else's pilot to finish.

Alongside Agent+, our team also builds custom enterprise AI solutions tailored to your exact compliance and operational needs.

Learn more at otherworldsai.com