The Bandwidth: Decoding Health + AI

Pope Leo XIV's Magnifica Humanitas (outlining the papacy's positions on AI) received significant coverage when it dropped on May 25. Most of the coverage focused on the encyclical's broadest arguments: AI and human dignity, the threat to democracy, the critique of autonomous weapons. What got far less attention were the passages focused on health, some of which are among the document's most pointed and specific. So let's dive in.

What the Encyclical Actually Says About Health

In addition to a general concern about people's mental health in an age of great digital disruption, the document makes several health-specific arguments.

On access to medical advances, the encyclical argues that scientific and technological progress in medicine "is not easily accessible to the vast majority of people." It describes wealthy regions spending on "superfluous interventions or dreams of individual enhancement accessible only to a select few" while other parts of the world lack basic equipment needed to save lives.

On AI in clinical decision-making, the document argues that automated systems making sensitive decisions in healthcare risk creating new forms of exclusion because they lack "compassion, mercy, forgiveness." No algorithm, it argues, can substitute for the moral judgment that healthcare requires. And in a passage on suffering and aging, it insists that illness, frailty, and vulnerability are not defects to be engineered away but dimensions of human experience through which compassion and wisdom grow.

The document's most striking healthcare passage addresses health data. It warns that entire regions are being subjected to the extraction of "health data, epidemiological profiles, genetic maps, and demographic information," which it calls "the new 'rare earths' of power." The pope argues that "Those who control the health data of entire peoples — often collected under the pretext of aid, research or innovation — possess a structural leverage over the future, for they can shape needs and markets." The encyclical calls this "one of the most urgent moral challenges of our time."

We Design This. We Decide.

If you follow this newsletter closely, you know that I have written about philosopher and ethicist Carissa Véliz, who reminds us that technology is not God-given. We are designing it, and we can design it differently. How much power these systems have, what goes into them, what decisions we allow them to shape — these are choices, not inevitabilities.

The encyclical is, in effect, a 42,000-word elaboration of that same premise. Technology is never neutral, Pope Leo XIV writes, because it "takes on the characteristics of those who devise, finance, regulate, and use it." The document does not argue that AI is inherently dangerous. It argues that AI is inherently shaped by human values and that those values need to be contested, named, and held accountable.

Where I'm Less Worried: Dehumanization

One concern raised in the encyclical (and underlined this week by AMA CEO Dr. John Whyte in a LinkedIn post) is that AI will progressively strip healthcare of its human character. Whyte's LinkedIn post says this: "The greatest risk in the AI era is not that machines become more intelligent. It is that healthcare becomes less human." At this moment, I am less persuaded that this particular outcome is the one we should be most focused on, because the evidence keeps pointing in a different direction.

I have written in this space before about what nurses and physicians describe as the art of medicine. An ICU nurse at Kaiser Permanente previously noted: "Sometimes you can see a patient and, just looking at them, know they're not doing well. It doesn't show in the labs, and it doesn't show on the monitor. We have five senses, and computers only get input."

And there was a story just this week about a woman who left her therapist after discovering sessions were being recorded by an AI scribe without her consent. Trust was broken, and could not be restored. Furthermore, as the story noted, “the presence of AI changes the therapeutic experience” by essentially introducing a third party into therapy sessions and raising concerns about privacy in a setting where absolute privacy is a must.

For now, the human-ness of medicine appears to be remarkably resistant to being algorithmized – if that's a word. I think it will continue to be. Mostly because we humans (as patients or providers) will demand it.

DeepMind CEO: AI Agents Are a "Practice Run" for AGI — Pope Leo was not the only one issuing warnings this past week. Speaking at Google I/O, DeepMind CEO Demis Hassabis said the current agentic era should be understood as a societal stress test for far more powerful systems to come, with AGI potentially arriving as early as 2029. He warned that governments and economists have only a few years to prepare — and that most aren't taking the timeline seriously enough.

An AI Biotech CEO Sets the Record Straight on AI Drug Development Hype — BigHat Biosciences CEO Peyton Greenside pushes back on speed claims dominating AI drug development: designing a molecule may take hours, but validating it still requires months of downstream testing that can't be skipped. Her argument is that the real opportunity for AI in biotech isn't faster de novo design, but more sophisticated molecular engineering that competitors — including Chinese pipelines — can't replicate.

AI Leaders Are Reshaping the Health System C-Suite — Health systems from HCA to Mayo Clinic are formalizing AI leadership roles, and the job description has matured: where early hires were expected to build momentum without a blueprint, organizations now have clearer priorities and are looking for executives who can govern, scale, and measure AI across clinical, IT, and legal functions. The share of healthcare organizations deploying AI grew from 3% to 22% between 2023 and 2025 — and with that growth comes pressure to answer harder questions about oversight and value creation.

How Health Systems Divide Claim Denials Work Between AI and Humans — With insurer denials claiming 6.6% of net revenue from commercially insured inpatient care in 2025 — up from 5.4% the prior year — health systems are deploying AI to handle routine cases and proactively flag denial patterns, while retraining staff to shift from responding to denials to preventing them.

Rural Healthcare Can't Wait for AI — Here's How We're Taking Action Now — Essentia Health's CMIO and director of data science describe deploying AI to more than 3,000 healthcare workers across a predominantly rural service area in Minnesota, North Dakota, and Wisconsin — focusing on EHR-embedded tools that govern centrally and reduce documentation burden without requiring behavior change on top of learning new technology. Their practical advice: don't chase one complex use case; identify five to ten high-volume, high-value workflows and implement broadly.

OpenAI Launches Biodefense Program — OpenAI announced the Rosalind Biodefense Program, offering its GPT-Rosalind model to trusted developers building biodefense and pandemic preparedness tools — spanning epidemiological modeling, early detection, non-pharmaceutical interventions, and medical countermeasure development. The company has briefed the White House and several federal agencies and is expanding access to U.S. government and allied partners for public health and biodefense missions.

Zuckerberg's Philanthropic Venture Unveils AI World Model for Drug Discovery — Biohub, the Chan Zuckerberg Initiative's biomedical research arm, launched an open-source AI world model designed to accelerate drug discovery by learning from protein sequences. Researchers used the models to design protein binders for cancer and immune targets that successfully reactivate immune cells in lab tests.

Inside Optum Health's Push to Make AI Practical for Clinicians — Optum Health recently expanded an AI-driven chart summarization tool to thousands of providers after a phased rollout that included a two-month regional pilot, clinical validation, and predetermined performance thresholds before broader deployment. Early clinician feedback points to reduced after-hours administrative work and smoother visit preparation — though all reported findings come from Optum Health's own internal assessments.

Health Leaders Talk How AI Can Help Patients Be More Proactive — At the TIME100 AI Leadership Forum, Rush University CEO Dr. Omar Lateef argued that AI's biggest near-term impact on health equity may be in medication adherence rather than new therapeutics — citing a pilot where an AI communication tool helped 20% of patients with uncontrolled hypertension bring it under control. Panelists converged on a broader thesis: AI's value in health isn't replacing clinical judgment, but translating data into action and closing the gap between what patients know they should do and what they actually do.

Introducing Claude Opus 4.8 — Anthropic launched Claude Opus 4.8 with improvements in agentic reliability and a notable honesty gain: the model is approximately four times less likely than its predecessor to let flaws in its own code pass unremarked. The release was paired with a $65B fundraise that takes Anthropic's valuation to $965B — surpassing OpenAI — and a commitment to bring Mythos-class models to general availability in the coming weeks.

Apple's Revamped Siri, Built on Google Gemini, Takes Shape — Bloomberg reports that Apple's long-delayed AI overhaul is finally taking form in iOS 27: a rebuilt Siri powered by Google Gemini, with a swipe-down interface for AI search, chat, and iOS tasks, and a dedicated ChatGPT-style app with support for third-party AI agents. Apple has been slow in the AI race, promising features in 2024 that never shipped, while more than a billion iPhone users wait for AI to arrive on the device they use every day.

Doctors, This Is Why Our Patients Are Using ChatGPT — Columbia ER physician Helen Ouyang writes that her own experience using ChatGPT for a routine health concern revealed something the medical system struggles to reliably offer: sustained attentiveness, patience, and personalized follow-through. Her conclusion is pointed — AI won't replace doctors, but it is already reshaping what patients expect from them.

What Have We Learned from Doctronic's AI Experiment in Utah? — Five months into Utah's pilot allowing an AI chatbot to renew prescriptions for roughly 200 medications, early data shows the AI grants renewals 72% of the time with 91% physician agreement — though all review is currently conducted by Doctronic's own team, not independent auditors. Experts are urging caution: as one health system AI chief put it, a constrained AI processing selected refill requests under supervision is not the same as evidence that AI prescribing is safe or equitable.

How Stanford Patients Help Expose 'Fault Lines' in Health AI Adoption — Stanford Health Care has spent the last 18 months running patient panels to review AI tools before deployment, and the process has surfaced a recurring tension: health systems optimize against false positives (errors that look embarrassing), while patients care more about false negatives (things being missed). The panels have reshaped tool design in concrete ways, but the C-suite isn't obligated to act on the feedback — and at least one case revealed that by the time patients were consulted, the vendor contract had already been signed.

The Hidden Layer Every Healthcare AI Solution Is Missing — Most healthcare AI products lack a validated clinical knowledge layer beneath the model — and without it, outputs vary by context, diagnoses migrate unsupported into problem lists, and coded data ends up too vague to support quality reporting or audits. The piece argues health system leaders should pressure vendors on a core set of questions: What clinical knowledge sits underneath the model, who curates it, and can the system explain why it reached a particular conclusion?

Health AI Chatbots Are Legally Medical Devices; It's Time the FDA Started Treating Them Like It — Two Harvard Law student-researchers argue that ChatGPT Health meets the statutory definition of a medical device under the FDCA and that OpenAI's disclaimers about intended use aren't enough to escape FDA jurisdiction — pointing to the agency's broad authority to consider actual consumer use patterns, product design, and distribution context. With Congress stalled on comprehensive AI legislation and state laws facing federal challenge, they argue the FDA is the last line of regulatory defense.

States Continue Efforts to Regulate AI in Healthcare: A Review of Legislation Passed in 2026 — A recent Holland & Knight review of 2026 state legislation finds a consistent through-line across diverse approaches: AI may assist insurers in prior authorization decisions, but cannot serve as the sole basis for denying care. Beyond insurers, states are also moving to restrict autonomous clinical decision-making by providers, mandate patient consent for AI tools in behavioral health settings, and impose disclosure and safety requirements on AI chatbots — particularly where minors are involved.

The House Democrat Taking a Big Risk to Land an AI Deal — Rep. Lori Trahan (D-MA) is quietly pursuing bipartisan AI talks with Rep. Jay Obernolte (R-CA) without the explicit blessing of Democratic leadership, which is running its own AI framework process ahead of the midterms - underscoring how fractured congressional Democrats remain on AI regulation.

'It Isn't Canceled': Inside the White House Divisions on AI — Trump's near-signed AI executive order — which would have created a voluntary framework for AI companies to give the government advance review of new models — was derailed at the last minute by former AI czar David Sacks, who called the president directly to flag industry concerns. Politico reports three distinct camps have emerged in the West Wing: a deregulatory faction led by Sacks, a hawkish bloc around Defense Secretary Hegseth worried about national security risks, and a middle-ground camp around Chief of Staff Wiles pushing for voluntary safeguards.

Roles and Responsibilities: Threshold Questions in Enterprise AI Adoption — As AI moves from pilot projects into core corporate workflows, boards face unresolved questions about privilege, record creation, and accountability that existing legal frameworks weren't built to answer. A Wachtell Lipton memo published via the Harvard Law School Forum on Corporate Governance outlines four threshold considerations for executives and boards — and makes clear that deploying AI in roles once reserved for humans requires explicit lines of human responsibility assigned in advance.

Economic Futures in the Age of AI — The OpenAI Foundation announced a $250M commitment to building economic security in the AI era, focused on three areas: measuring how AI shifts value across workers, firms, and capital owners; supporting workers through near-term disruption; and developing new mechanisms — including sovereign wealth fund models and adaptive tax structures — to ensure broad-based sharing of AI's economic gains.

Integration of Artificial Intelligence Into a Medical Curriculum: Evolving Student Perceptions and Faculty Development Challenges — A peer-reviewed study from East Tennessee State University tracked medical student and faculty attitudes toward AI across two years, finding that students are increasingly using AI for active learning tasks like practice question generation and self-testing, while consistently ranking the ability to function without AI as their top professional priority. Faculty confidence improved after a skills-focused workshop but declined after one centered on ethics and risk — a pattern the authors interpret as a shift from “unconscious incompetence to a more realistic, if more measured, understanding of the technology.”

Impact of an Ambient AI Scribe on Medical Student Clinical Examination Notes — A nonrandomized clinical trial of 104 first-year medical students found that hybrid notes — created by revising student-written notes with AI-generated content — showed minimal differences in overall quality compared to human-only notes, with the greatest benefit accruing to lower-performing students. Students found AI notes useful as a starting point but flagged frequent omissions of key details, and a subset worried that AI reliance could erode the foundational documentation and reasoning skills they're there to develop.

United States AI Adoption Shows Steady Growth, but Distribution Remains Uneven — A new Microsoft report finds that more than 30% of the U.S. working-age population now uses AI — up three percentage points since the end of 2025 — but usage in metropolitan counties runs at nearly double the rate of rural areas (32.9% vs. 16.2%), underscoring a geographic divide that mirrors broader digital equity gaps. Despite leading the world in AI innovation, the U.S. ranks just 21st in global AI adoption.

How AI Is Forcing McKinsey and Its Peers to Rethink Pricing — AI is accelerating a shift toward outcome-based pricing across professional services, as clients question the value of advice billed by the hour when consultants are using AI to do much of the analytical work. McKinsey is responding by tying more partner compensation to equity and restructuring how fees are set — a signal that the billable-hour model has a shorter runway than many assumed.

Managers Are Struggling to Keep Up with the AI Productivity Boom — As AI collapses execution timelines, managers are becoming the new organizational bottleneck — overwhelmed by the volume of decisions, reviews, and feedback that now arrive faster than traditional management systems were designed to handle. The authors argue the fix isn't slowing teams down but shifting the manager's role from editor-in-chief to strategic guide: setting direction, clarifying priorities, and reserving human judgment for the decisions that actually require it.

Coalition for Health AI Unveils Governance Playbooks for Responsible AI Adoption — The Coalition for Health AI released eight governance playbooks covering areas from organizational AI policy and risk assessment to third-party management and staff training — developed through collaboration with more than 100 healthcare organizations and designed to give health systems a practical, adaptable framework for responsible AI implementation. The playbooks align with the Joint Commission's planned voluntary AI certification program, offering a common operating language for governance regardless of an organization's size or resource level.

Tokenmaxxing Is Over. That's Because It Never Measured What Really Counts to see ROI from AI — Companies that rewarded employees for maximizing AI token usage ran straight into Goodhart's Law: the metric became the goal, producing meaningless busywork at Amazon and eye-watering bills at Uber and Salesforce. The deeper problem, Jeremy Kahn at Fortune argues, is that most firms are stuck in early-stage AI adoption — replacing processes rather than reinventing them — and won't see real productivity gains until they redesign entire business models, not just workflows.

👉 Jun. 1–5 — Stanford Health AI Week

👉 June 2 — CHAI Webinar: The Cost of Catching Up — Why AI Governance Can't Wait Until Deployment — Virtual

👉 Jun. 8–10 — Fortune Brainstorm Tech, Aspen, CO

👉 Jun. 9–18, 2026 — Google Cloud-Hosted Startup School: Agentic AI — Virtual

👉 Jun. 15–18 — Databricks Data + AI Summit 2026, San Francisco + Virtual

👉 June 24 — CTRC 14th Annual Digital Health Summit: Bridging the Gap — Digital Health and AI for Smarter Care, Stronger Teams, and Sustainable Healthcare — Virtual

👉 Jul. 7–10 — AI for Good Global Summit 2026, Geneva, Switzerland

👉 Aug. 4–6 — Ai4, Las Vegas

👉 Oct. 13–16, 2026 — AIxPH 2026: 1st Annual Conference on Artificial Intelligence and Public Health, Baltimore, MD (Abstract submissions due May 29)

And finally, if you like what you are reading, please share this newsletter with your networks and encourage them to sign up. ✍️ 🆙 And/or, give me a shout out on LinkedIn.

Till next time,

BC