NHS England is withdrawing open-source software from public repositories in response to new threats from AI models built to help cybercriminals. The decision has sparked swift pushback from developers, clinicians, and digital rights groups who warn the change will weaken transparency, slow innovation, and fail to improve security.
The move comes as so-called “computer-hacking AI models,” including tools described as Mythos, gain attention for automating parts of cyberattacks. Officials are acting to limit what they see as exposure of sensitive code. Critics argue the risk is overstated and that the benefits of open code—peer review, faster fixes, and reuse—are being lost at a sensitive time for the health service.
Background: Why Open Code Matters in Health Tech
Open-source code has supported parts of the NHS digital stack for years. It allows outside experts to spot bugs, suggest patches, and reuse proven components. That model can lower costs and make systems easier to audit. Health software also faces strict safety and privacy demands, which benefit from public scrutiny.
Security concerns have risen as attackers shift tactics. AI-assisted tools can scan code, map systems, and help write exploits. Some of these capabilities are now packaged so that less-skilled actors can attempt sophisticated attacks.
“NHS England is pulling its open-source software from the internet because of fears around computer-hacking AI models like Mythos.”
The policy change suggests leaders believe public code increases the attack surface. Opponents say the threat is not the code itself but unpatched systems and weak controls.
Reactions From the Digital Community
Developers and data leaders voiced concern that taking code private will reduce oversight. They say public repositories act as an early warning system, with many eyes reviewing changes. Closed code can slow updates and hide issues until they become incidents.
“Opposition is growing among those who say the move is bad for transparency and efficiency, and will also do nothing to improve security.”
Some clinicians who rely on digital tools worry about delays to fixes and features. Patient safety advocates add that auditability helps build trust in software used for care and logistics.
The Security Debate: Exposure Versus Resilience
Security experts are split. One camp argues that public code can reveal implementation details that aid attackers. They support limiting access while threat models evolve. The other camp points to “security through obscurity” as a weak defense. They say strong configuration, logging, rapid patching, and zero-trust designs matter more than hiding code.
There is also a practical point. Attackers often reverse-engineer binaries or probe live systems. Even private code can leak through dependencies, vendor builds, or misconfigurations. Open models can accelerate both attack and defense. Taking code offline may slow defenders more than attackers.
- Proponents of removal: reduce reconnaissance data for AI tools.
- Opponents: keep peer review and faster patch cycles alive.
- Middle ground: restrict sensitive modules, keep non-sensitive code open.
Impact on Costs, Delivery, and Collaboration
Shifting to closed repositories can increase vendor lock-in. Public code lets trusts adapt tools without lengthy contracts. It also supports shared solutions across regions. If code goes private, teams may duplicate work and pay more for similar outcomes.
Procurement timelines could lengthen as each change requires approvals. That can slow clinical improvements tied to digital workflows, from scheduling to supply tracking. Universities and small firms that contribute fixes may step back without visibility into core projects.
What Happens Next
Several groups are pushing for a review period. They want risk assessments that separate high-risk modules from low-risk utilities. They also call for red-team exercises, bug-bounty programs, and clearer patch SLAs. These measures aim to harden systems without losing the advantages of open collaboration.
Independent auditors suggest a compromise. Keep code public where safe, but remove credentials, secrets, and infrastructure maps. Use private mirrors for sensitive integration layers. Invest in continuous scanning, dependency management, and rapid rollout of fixes.
NHS England’s decision highlights a larger question for public bodies under threat from AI-enabled attacks. Should they hide more, or build stronger processes in the open? The outcome will shape how the health service balances speed, security, and trust. Watch for updated guidance, a possible phased approach to code access, and new incentives for responsible disclosure. The next few months will show whether privacy of code brings fewer incidents—or whether transparency remains the stronger shield.
