In partnership with

AI Spotlight — Washington's Big AI Health Experiment
AI SPOTLIGHT

Washington's Big AI Health Experiment

OpenAI and Anthropic are handing 2,000 public health workers enterprise AI licences. What could possibly be undefined about that?

📖 6 minute read
Public health worker reviewing data on a laptop in an office setting

Welcome Back,

Local health departments have spent years being asked to do more with fewer people and older systems. Now two of the biggest names in AI are stepping in to see whether that can change.

OpenAI and Anthropic, working alongside the Coalition for Health AI and Accenture, are launching a pilot programme that puts enterprise AI tools directly into the hands of public health practitioners across ten U.S. jurisdictions. It's called PULSE, and it's meant to answer a question nobody has tested at this scale before, whether generative AI actually belongs inside the messy, high-stakes world of public health work.

Today we look at what PULSE actually covers, why nearly 40 percent of local health departments are sitting out AI entirely, and the long list of governance questions the announcement leaves wide open.

📌 In Today's AI Spotlight

  • What PULSE is and who's actually running it.
  • The five public health use cases practitioners will test.
  • Why nearly 40% of local health departments still use no AI at all.
  • The governance and privacy questions the announcement leaves unanswered.
  • Our AI Spotlight take on what this pilot really tests.

🏥 What Exactly Is PULSE

PULSE stands for the Public Health Use Case and Learning Scaling Engine, a programme built by the Coalition for Health AI (CHAI) with OpenAI, Anthropic, and Accenture as partners, according to AI News.

The setup is straightforward on paper. OpenAI and Anthropic are each donating ten enterprise licences, together covering up to 2,000 public health practitioners across ten state, local, tribal, or territorial jurisdictions. Accenture handles onboarding and helps turn the results into implementation playbooks other agencies can borrow from later.

"Every major technological transformation succeeds or fails based on trust, governance and execution."

— Dr. David Lakey, former Texas health commissioner

Notably, CHAI hasn't disclosed which specific products, model versions, or configurations will actually be deployed, or how the two providers will be split across the ten pilot sites. The pilots are set to begin this autumn, with playbooks expected in 2027.

Healthcare data displayed on a screen with charts and maps

PULSE spans ten jurisdictions, from state health departments to tribal health authorities, testing AI in real operational settings rather than labs.

🔍 Five Use Cases, One Big Test

CHAI's leadership council will assign practitioners to communities working on five specific applications, spanning biosurveillance and drug-wave prediction, mapping social determinants of health, operational efficiency and community-feedback analysis, multilingual public communications, and automated clinical-data retrieval through a FHIR query engine, per AI News.

That FHIR piece is worth pausing on. FHIR is the healthcare industry's standard for exchanging patient data electronically between systems, and the announcement doesn't spell out exactly what role the AI plays there, whether it's writing the queries, retrieving the records, summarizing what comes back, or doing all three at once.

💡 AI Spotlight Take

Automating clinical-data retrieval sounds efficient right up until you ask who checks the AI's summary before a nurse or epidemiologist acts on it. That detail isn't in the announcement yet, and it's probably the most important one.

The other use cases carry similar open questions. Biosurveillance and drug-wave prediction touch public safety decisions directly, while the translation hub and community-feedback tools shape how agencies communicate with residents during health emergencies.

Granola Runs Revenue On Attio

"When I think of revenue, I think of Attio." - Shreman Shrestha, Head of Business at Granola

Here's what that adds up to:

  • Zero missed leads and 10x faster access to customer context

  • Lead triage 83% faster

  • Five hours saved per week with automated updates

AI Spotlight — Washington's Big AI Health Experiment Part 2

📊 Why Public Health Is Starting From Behind

Part of the case for PULSE is how far behind public health agencies already are. Data cited by CHAI from the National Association of County and City Health Officials found that nearly 40% of local health departments aren't using AI in any form today.

PULSE By the Numbers

10

jurisdictions selected to run pilot deployments

 

2,000

public health practitioners with licence access

 

~40%

of local health departments use no AI today

Dr. Brian Anderson, CHAI's chief executive, has pointed to how public health agencies entered the COVID-19 pandemic after years of thin technology investment, framing PULSE as a way to build practical AI experience before the next major public health event arrives, rather than during one.

Team of health workers reviewing documents and a laptop together

CHAI says agencies are interested in using AI to revise workflows and improve operational efficiency, not just add new tools on top of old ones.

🔒 The Governance Questions Still Open

Here's where the announcement gets thinner. CHAI hasn't published separate evaluation, privacy, security, or human-review requirements for any of the five use cases, and it hasn't said whether the pilots will work with identifiable patient records, de-identified data, synthetic data, or aggregated datasets.

HIPAA, the federal law protecting health information, will apply unevenly across the programme, its reach depends on which agency, which data, and which specific workflow is involved, not on the pilot as a whole. OpenAI and Anthropic both state that business-tier products aren't used to train models by default, but that policy alone doesn't define retention periods, access controls, or audit requirements for PULSE deployments specifically.

"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."

— Elizabeth Kelly, Anthropic's head of beneficial deployments

The announcement is also quiet on human oversight, it doesn't specify whether staff must sign off on generated public communications, verify AI translations, or validate clinical data before any of it gets used operationally. NIST's AI guidance recommends exactly that kind of review structure, but PULSE hasn't confirmed it will be built in from day one.

⚖️ What's Actually Defined vs. What Isn't

Reading through the announcement, there's a clear split between the logistics CHAI has nailed down and the safeguards it hasn't yet.

Still Undefined in the PULSE Announcement

⚠️  Which models and configurations each jurisdiction will actually use
⚠️  Whether pilots will touch identifiable, de-identified, or synthetic patient data
⚠️  Retention periods, access controls, and audit arrangements for each site
⚠️  Whether staff must review AI outputs before they're used operationally

None of this means the programme is reckless, pilots are often designed to work these details out as they go. But it does mean the "guardrails" language in the announcement is currently more aspiration than specification.

Person reviewing medical charts and data on multiple screens

The gap between "AI is being piloted" and "AI is being governed" is exactly what PULSE's 2027 playbooks are meant to close.

🧠 AI Spotlight Analysis

PULSE is a genuinely useful idea wrapped in a genuinely incomplete announcement. Handing enterprise licences to 2,000 practitioners is a real, concrete step, and testing AI in actual public health workflows beats testing it in a vendor demo. The problem is that "governance and responsible use built in from the start," as Anthropic put it, is currently a promise rather than a published set of rules.

Dr. Ashish Jha, a former White House COVID-19 response coordinator, framed the stakes plainly, arguing the real question isn't whether AI reshapes public health, it's whether that happens thoughtfully.

💬 Quote of the Week

"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

The honest read is that PULSE is a test of two things at once, whether these AI tools genuinely help under-resourced health agencies, and whether the industry can build governance in real time instead of after something goes wrong. The 2027 playbooks will tell us which one won.

💡 Final Thoughts

PULSE isn't trying to replace public health workers, it's trying to give chronically under-resourced agencies a real shot at AI tools they've mostly sat out until now. That's worth taking seriously.

But the gap between "we built in privacy and governance from the start" and an actual published rulebook is still wide open. Whether PULSE becomes a genuine model for responsible public-sector AI or a cautionary tale will come down to what gets nailed down before autumn, not after.

Would you trust an AI-drafted clinical summary or public health alert before a human reviewed it? Hit reply, we read every response.

🔗 Sources and Further Reading

AI News: US public health agencies to test OpenAI and Anthropic AI models

❤️ Enjoying AI Spotlight?

If today's edition helped you see the tradeoffs in public-sector AI more clearly, consider sharing it with a colleague, founder, or friend interested in technology.

Share AI Spotlight →

Thanks for reading AI Spotlight.

Our mission is simple: deliver clear, trustworthy, and actionable AI insights that help professionals stay ahead without the hype.

Keep Reading