Oracle's Healthcare Data Monopoly: The Contract Nobody Can Find
One company holds the clinical records and the CMS contract covering 150 million Americans — and HIPAA enforcement just lost 20,000 of the people supposed to protect that data.
Oracle now occupies a position in American healthcare data infrastructure that should alarm every state legislator who votes on health policy. Through its acquisition of Cerner and a Centers for Medicare & Medicaid Services contract covering more than 150 million Americans, Oracle holds both the clinical records and the billing pipeline for a majority of the country's healthcare interactions.
The CMS contract itself raises immediate governance questions. It does not appear on SAM.gov — the federal government's official procurement database — which means there's no visible evidence of competitive bidding, no public record of the contract terms, and no straightforward way for state officials or oversight bodies to review what Oracle agreed to do and under what constraints.
87%of Americans can be re-identified using just three data fields — ZIP code, birth date, and sex — that HIPAA does not require removal from datasets. Oracle now controls the infrastructure where these records live.
Meanwhile, the Department of Health and Human Services has shed approximately 20,000 employees in recent months, gutting the enforcement capacity that HIPAA depends on. The law was already inadequate — designed for a world of paper records and fax machines, not one where a single cloud provider can cross-reference clinical data with billing records at scale. Without enforcement staff, HIPAA becomes a suggestion.
For state health officials: your Medicaid data flows through Oracle's infrastructure. Your hospitals run on Cerner. Your enforcement backstop at HHS has been hollowed out. If your state hasn't conducted an independent review of what data Oracle holds on your constituents and under what terms, that review is overdue. The federal government is not watching. Someone needs to be.
Source: Drey Dossier investigation · YouTube intelligence brief, March 2026
Feature
The Pre-Maxwell Phase: Why AI Still Can't Unify Its Own Skills
Source: Discover AI · Physicist analysis, March 2026
A theoretical physicist argued this month that artificial intelligence is in its "pre-Maxwell phase" — a reference to the 19th century moment when electricity and magnetism were understood as separate phenomena before James Clerk Maxwell proved they were aspects of the same underlying field.
The analogy is precise. We've built language models that reason about text. We've built vision models that parse images. We've built tool-using agents that execute code. But nobody has found the unified theory that explains why these capabilities emerge, how they relate, or what the deeper structure is. We're doing AI the way 1850s physicists did electromagnetism: cataloging phenomena without understanding the field.
The current approach is atomic skill fragmentation — benchmark each capability in isolation, optimize each one separately, and hope coherence emerges from the stack. It hasn't.
What makes this more than academic is a detail from Google's own research. When Gemini was asked to characterize its own emergent capabilities, it described them as existing "in the space between the skills" — not in any individual benchmark, but in the transitions and relationships between tasks. This matches the Structured Emergence research framework: that meaningful AI development happens through interaction and relationship, not through isolated capability gains.
The policy implication is significant. Every AI safety framework currently in use evaluates models based on discrete capabilities — can it generate code, can it reason about biology, can it manipulate users. If the real risk (and the real capability) lives in the spaces between those skills, then our evaluation frameworks are looking at the wrong thing. We're measuring the electricity and the magnetism separately while the electromagnetic field goes unmonitored.
Data Point — AI Job Exposure by Occupation (Karpathy Index)
Andrej Karpathy's AI job exposure scale (0–10). Average across all occupations: 5.3. Higher means more of the job can be performed or replaced by current AI systems.
Software Devs
8.5
Copywriters
9.0
Paralegals
8.0
Tax Preparers
8.5
Avg (All Jobs)
5.3
Nurses
3.0
Electricians
1.5
Roofers
0.5
Source: Andrej Karpathy · March 2026 · Scale: 0 (no exposure) to 10 (fully automatable)
Oklahoma Focus
Well Repurposing Act (HB 3173) Passes House
Rep. Nick Archer (R-Elk City) · Passed House March 16, 2026 · KGOU/StateImpact
The Well Repurposing Act passed the full Oklahoma House floor on March 16 after clearing the Energy Committee unanimously (8-0) on February 6. The bill authorizes conversion of abandoned and orphaned oil and gas wells for geothermal energy production and energy storage at or above 250°F, with oversight by the Corporation Commission or the Department of Environmental Quality.
Oklahoma has more than 19,000 orphaned wells with unreachable or unknown owners. According to OCC Chief Financial Officer Holly George, at current plugging rates it would take 235 years to address the backlog. As Rep. Archer put it: “If there’s an opportunity to avoid some of those costs, utilizing the existing infrastructure and repurposing it, that may be economically viable to private capital.” The bill is modeled after New Mexico legislation signed into law in 2025.
Next: Heads to Oklahoma Senate. If enacted, takes effect November 1, 2026.
Advancing — In Senate
OU Geothermal Permit: First in State, Now Paused
University of Oklahoma · Wells of Opportunity Project, Tuttle
The University of Oklahoma's "Wells of Opportunity" project, directed by Jeff McCaskill, secured the first permit in Oklahoma to convert oil and gas wells for geothermal purposes. The pilot at Tuttle aims to pipe geothermal heat from donated wells — about 1.5 miles south of Tuttle Elementary and Middle Schools — to offset school utility costs, backed by $1.7 million in DOE funding as part of an $8.4 million federal initiative. The wells were donated by Blue Cedar Energy, managed by OU alumni.
The project is currently paused, halted by the Trump administration and awaiting DOE approval for its next phase. McCaskill has flagged a concern with HB 3173’s 250°F temperature threshold, arguing it should be lowered: “Drop that temperature down to 180 degrees or something … makes it feasible for people to start saying, ‘I can convert one of these wells over.’” The gap between the bill's threshold and the project's practical needs illustrates the regulatory fine-tuning still ahead.
Paused — Awaiting Federal Approval
Chief AI Officer Tai Phan: "Accountable Innovation"
Adobe Government Forum, Washington D.C. · February 2026 · Route Fifty profile
Oklahoma's first Chief AI and Technology Officer spoke at the Adobe Government Forum in February, rebranding the state's AI approach with a pointed redefinition: “AI for us, moving forward, stands for accountable innovation.” Key positions: AI systems must be transparent about data sources and training methods, the state's focus is amplifying workforce capability rather than reducing headcount, and IT leadership must be involved earlier in procurement decisions to improve return on investment.
Phan has also pushed for process redesign alongside technology adoption and building leader competency in evaluating emerging technologies — arguing that reliability depends on qualified personnel, not just tools. Appointed November 2025 through OMES at $190K with 15+ years in retail, health, and technology, he's positioned Oklahoma as an early-mover state in AI governance — an unusual stance for a state more commonly associated with oil and agriculture.
Active — Framework Development
The Thread
Every story in this issue is about the same thing: who controls the infrastructure that everyone depends on.
Oracle controls the healthcare data infrastructure for 150 million Americans, with no visible public contract and an enforcement agency that's been gutted. The Pre-Maxwell problem means we're evaluating AI capabilities in fragments while the real power lives in the spaces between skills — spaces nobody's governing. Oklahoma is trying to reclaim physical infrastructure — turning abandoned wells into geothermal assets — while simultaneously building governance infrastructure through its AI office.
The Karpathy numbers make the stakes personal: a 5.3 average means most people's jobs are already partially exposed, and the most exposed occupations are exactly the knowledge workers who write policy, draft contracts, and review compliance.
The question for every state: are you building the governance capacity to match the speed of consolidation? Or are you waiting for someone else to do it?
The Data
The Widening Gap
AI investment is accelerating exponentially. Regulatory response remains fractured and reactive. The distance between them grows every year.
Sources: Crunchbase AI funding data, 2020–2026; PitchBook venture capital reports; MultiState AI legislation tracker (March 2026); White House OSTP archives; European Commission AI Act timeline. Investment figures represent estimated total private AI sector investment per year. | Design: The Inference