Backlash builds over NHS plan to hide source code from AI hacking risk
New Scientist · rss · 2026-04-20 Reader 🖥️
NHS England's decision to withdraw publicly funded open-source code from public access — motivated by fear of AI-powered vulnerability discovery tools like Mythos — is counterproductive because it removes the security benefits of open scrutiny while doing nothing to stop determined attackers, and it betrays the transparency obligations owed to taxpayers.
practical · argumentative — A news article structured as a policy critique — it reports on NHS England's decision to close-source its software, then marshals expert voices and institutional evidence to argue the decision is wrong. Every quoted source opposes the policy; the article is advocacy journalism in news format.
Structure
| Part | Role |
|---|---|
| The policy announcement | Context — establishes the triggering event: NHS England ordering all open-source code pulled from public repositories by 11 May due to Mythos risk. |
| The open letter and public figures | Social proof — shows breadth of opposition (682 signatures, Cory Doctorow, Matt Hancock) to establish the backlash is substantial and cross-partisan. |
| Hancock's argument: taxpayer value and security through openness | Evidence — articulates the positive case for open source: taxpayer ownership, more rigorous testing, global contribution. |
| Harbuz's technical rebuttal | Expert testimony — a researcher who actually used Mythos against NHS code explains that the real problem is chronic underinvestment in cybersecurity, not code visibility. |
| AISI's evaluation of Mythos | Institutional authority — the UK government's own AI Security Institute found Mythos effective only against 'small, weakly defended and vulnerable' systems, undermining the threat model motivating the policy. |
| Terence Eden's transparency argument | Values framing — connects open source to public trust and democratic accountability in healthcare. |
| NHS England's response | Counterpoint (minimal) — the only defence offered is a boilerplate statement about 'temporarily restricting access' while assessing AI developments. |
Key terms
- Mythos Mythos
An AI model created by Anthropic reported to be capable of discovering software vulnerabilities at scale, cited as the proximate cause of NHS England's policy change. - open source (in NHS context) Public-Sector Digitalization
The NHS service standard requiring publicly funded software to be published openly so it can be built upon, improved, and reused — treated as both a transparency obligation and a security practice. - responsible disclosure AI-Driven Vulnerability Discovery
The process Harbuz's team followed: discovering vulnerabilities in NHS code using Mythos and reporting them privately to NHS before publication.
Propositions
- explicitWithdrawing source code from public view does not improve security because backups still exist, AI models have already been trained on the code, and conventional security tools can find the same vulnerabilities.conf: high
- explicitThe real security problem is systemic underinvestment in cybersecurity, which predates Mythos.conf: high
- explicitClosing the source code hurts security helpers and contributors more than it hurts attackers.conf: high
- explicitPublicly funded code belongs to the public and should be available to the public.conf: high
- explicitcontestedMythos is only effective against small, weakly defended, and already vulnerable systems — not well-secured infrastructure.conf: highcontradictsAssessing Claude Mythos Preview's cybersecurity capabilities · p1 — The claim that Mythos is only effective against small, weakly defended systems clashes with Anthropic's claim of zero-day exploits across every major OS and browser.
- explicitPublic trust in the NHS depends on transparency, and digital tools must therefore be open source.conf: medium
- implicitThe NHS's own vulnerability to Mythos is evidence of pre-existing security debt, not evidence that open source is dangerous.conf: high
- implicitSecurity through obscurity is an ineffective strategy against sophisticated AI-powered threat actors.conf: high
Arguments
NHS England should reverse its decision to close-source its code. strength: high
Explicit premises:
- Closing the code does not prevent attackers from finding vulnerabilities (code is already cached, AI models already trained on it)
- Closing the code does prevent security researchers and contributors from improving it
- The UK government's own AISI found Mythos is not effective against well-secured systems
- The NHS service standard requires open-source publication
Implicit premises:
- Security through obscurity is ineffective as a primary defence strategy
- The NHS's threat model should be based on AISI's assessment rather than media reporting about Mythos
- The policy was driven by fear and media coverage rather than rigorous threat analysis
Inferences:
- If Mythos found vulnerabilities in NHS code, the appropriate response is to fix the vulnerabilities, not hide the code
- The policy may create a worse security outcome than the status quo by removing community scrutiny while not actually reducing attacker capability
The vulnerabilities Harbuz's team found demonstrate the value of open source, not its danger. strength: medium
Explicit premises:
- The team responsibly disclosed the vulnerabilities to NHS
- The same vulnerabilities could be found by conventional tools
Implicit premises:
- Responsible disclosure by white-hat researchers is a net security benefit
- The counterfactual — those vulnerabilities existing undiscovered in closed code — would be worse
Inferences:
- The NHS may have used Harbuz's disclosure as justification for closing the code, which would be an ironic inversion — punishing the mechanism that revealed the problem
Narratives
- Harbuz's team used Mythos to scan NHS open-source code, found severe vulnerabilities, and responsibly disclosed them — only for the NHS to respond by pulling all code from public view.
→ Supports the proposition that the NHS is drawing the wrong conclusion from the evidence — treating the symptom (code visibility) rather than the cause (poor security practices). - Matt Hancock, former health secretary who championed NHS digitisation, publicly opposes the decision on LinkedIn.
→ Demonstrates cross-partisan opposition and suggests the policy contradicts the NHS's own institutional trajectory toward openness.
Questions addressed
- ✅ Does withdrawing open-source code improve NHS cybersecurity against AI-powered threats?
No — expert consensus in the article is that it does not, because the code is already available elsewhere and the real problem is underinvestment in security. - ✅ How capable is Mythos actually, and does it justify the NHS response?
AISI found Mythos effective only against already-weak systems; the threat does not justify hiding well-structured code. - ⭕ What should the NHS do instead?
Implied but not explicitly laid out — the article argues against the current policy but does not propose a concrete alternative security plan. - ⭕ Is there any legitimate security rationale for temporarily restricting code access while conducting a security review?
Not addressed — the article does not engage with the possibility that a temporary pause during remediation might be reasonable.
Gaps
- incompleteThe article quotes no cybersecurity expert who supports the NHS decision, even conditionally. A one-sided expert roster makes the argument persuasive but analytically weak — we don't hear the steelman of the NHS position.
Read next: NCSC or NHS Digital cybersecurity strategy documents; any published rationale from the NHS decision-makers. - incompleteNo discussion of what a proportionate response to AI-accelerated vulnerability discovery would look like. The article is effective at saying the policy is wrong but provides no constructive alternative beyond 'keep it open.'
Read next: SLSA framework, Sigstore/provenance attestation for government software, mandatory security review pipelines.closed byProject Glasswing: what Mythos showed us — Provides the constructive alternative the article lacked: the proportionate response to AI-accelerated vulnerability discovery is architectural hardening (front-of-application defenses, blast-radius isolation, instant rollout), not hiding code or faster patching. - uninformedThe article does not distinguish between different types of NHS code — some repositories may contain infrastructure configurations, deployment scripts, or internal tooling where the security calculus differs from application code.
Read next: NHS England's GitHub repositories; analysis of what categories of code were actually published. - incompleteThe broader policy question — how should governments respond when AI dramatically lowers the cost of finding vulnerabilities in public infrastructure code — is raised implicitly but not addressed. This is arguably the most important question the article touches.
Read next: CISA guidance on AI and software security; EU Cyber Resilience Act provisions; academic work on the security economics of open-source critical infrastructure.
Related documents
- NHS England rushes to hide software over AI hacking fears
- Three reasons to think that the Claude Mythos announcement from Anthropic was overblown
- AI Companies Aren't Evil. But They Are Reckless.
- Assessing Claude Mythos Preview's cybersecurity capabilities
- How governments and organizations are leveraging Google's AI breakthroughs for crisis resilience
- 16) Inga Strümke om hvor AI er på vei, den outsourcede problemløsningen og datamanifolden
- IBM and Red Hat Commit $5 Billion to Redefine the Future of Open Source in the AI Era
- Meta's new model is Muse Spark, and meta.ai chat has interesting new tools
- OpenAI Releases GPT-6 Astra for Coding and Computer Use
- Radar Trends to Watch: May 2026