#73. Claude Mythos and What it Means for Healthcare

This week’s newsletter looks at Claude Mythos and why Anthropic’s decision not to release its most powerful AI model matters for healthcare, particularly in the context of cyber risk, organisational readiness and patient safety. I also reflect on how AI is beginning to influence clinical judgement in practice, why that shift raises questions about governance and accountability, and what recent leadership appointments in Ireland’s health system signal about the next phase of reform and delivery.
Claude Mythos and What it Means for Healthcare
Anthropic’s decision not to release its most powerful model, Claude Mythos, is one of the more telling AI stories of the week, if not the year. In internal testing, the model demonstrated an ability to autonomously find and exploit previously unknown software vulnerabilities at a scale that even its creators judged too risky for public access. Rather than pushing ahead, Anthropic has restricted Mythos to a small group of technology and infrastructure providers through Project Glasswing, framing it explicitly as a defensive cybersecurity tool rather than a commercial product.
Anthropic has a history of using fear to market their products(1), but this one does appear to be legitimate, mainly down to their behaviour. Building a model and then choosing not to release it is an unusual move in a market that rewards speed and scale. That suggests the internal assessments crossed a threshold where reassurance through safeguards was no longer judged sufficient.
For healthcare, the relevance is immediate. Hospitals and health systems run dense estates of ageing software, bespoke integrations, and connected medical devices that are difficult to patch and often poorly understood end‑to‑end. Security specialists have been clear that tools like Mythos compress the time needed to discover exploitable flaws from months into hours. Faster discovery increases the likelihood of coordinated, multi‑site disruption, with consequences that extend beyond data loss into operational continuity and patient care.
Seen through that lens, the more pressing question for healthcare organisations is whether they are institutionally ready for a world where vulnerabilities surface faster than traditional processes can absorb. This exposes long‑standing issues that technology alone will not fix: incomplete visibility of live systems, heavy dependence on suppliers for patching decisions, and incident response plans that look good on paper but are rarely exercised under realistic conditions.
Read more: https://techcrunch.com/2026/04/07/anthropic-mythos-ai-model-preview-security/
(1) - https://www.bbc.co.uk/news/articles/cpqeng9d20go
What does AI in Healthcare mean for clinical judgement and expertise?
Another great article from my colleague Ronan Glynn, arguing that as AI systems move from experimentation into everyday workflows, medical expertise is shifting from the authority of the individual clinician to a form of collective intelligence.
That creates a tension that clinicians are left to manage. Some AI systems perform very well in specific tasks, which makes it hard to dismiss their recommendations. At the same time, many of these systems don’t explain their reasoning in a way that fits clinical norms. The clinician still carries the responsibility for the decision, even when the logic behind it comes from a model they can’t fully interrogate.
The point the article lands on is less about technology and more about how health systems respond. If AI is influencing judgement, then training, governance and accountability need to be clearer than they are today. Clinical judgement isn’t disappearing, but it is changing shape, and earning trust will depend on how deliberately organisations design the human side of that change.
Appointment of Secretary General of the Department of Health in Ireland
The Irish government has appointed Derek Tierney as Secretary General of the Department of Health, succeeding Robert Watt att at a pivotal moment for the system. Confirming the appointment, the Minister for Health, Jennifer Carroll MacNeill TD MacNeill, pointed to Tierney’s ability to turn major public investment into delivery. With Sláintecare progressing, Health Regions taking shape, and significant capacity and digital programmes under way, the role will be central to aligning reform, infrastructure and execution around a more integrated and sustainable model of care. Good luck, Derek!
‘I came back because I really care’ – HSE’s new boss Anne O’Connor
Anne O’Connor has said her decision to return as HSE chief executive was driven by a personal commitment to public healthcare and a belief that the system now needs to work differently. Taking up the role following Bernard Gloster’s retirement, she pointed to the scale of recent investment, the pressures from an ageing population and the need to focus on keeping people well to reduce demand on services. While noting some improvement in overall waiting times, she acknowledged the challenge of nearly one million people on waiting lists and a projected budget overspend after winter pressures, signalling that corrective measures will be required alongside ongoing reform to deliver better outcomes from existing resources.
Originally published on LinkedIn.