I was at HPN North in February. Two days of conversations with NHS leaders, and the same thing kept surfacing in different ways: Trusts are deploying AI tools for various use cases right from ambient voice, diagnostic support to patient flow automation, but without a defined process for reviewing the security of what they’re buying.

Clinical safety leads are being asked to sign off new products on timelines that don’t allow for proper technical review.

AI strategies are being approved at board level where nobody has been made accountable for the cyber security of the tools within them.

This is just the reality of working in systems under sustained pressure, where clinical and operational urgency consistently outpaces governance.

Research from the Healthcare Strategy Forum puts some numbers behind this

70% of NHS leaders plan to increase AI spend this year. That is the top investment category, ahead of workforce, ahead of cloud, ahead of everything else.

66% plan to increase cyber security spend in the same period.

Both numbers are going up but the gap between them isn’t closing,  if anything it’s widening.

The picture, honestly

Let’s start with what the NHS actually has going for it, because it’s more than most sectors. The DSP Toolkit, the NCSC’s Cyber Assessment Framework, the Data Security and Protection requirements. There is a governance infrastructure in place that other industries would envy.

Most boards have cyber security represented at executive level. And the CCIOs and clinical safety leads we talk to are not complacent , they absolutely want to get AI governance right.

But the way AI is currently being bought and deployed has opened up a gap that the governance infrastructure hasn’t kept pace with. Clinical and operational priorities drive adoption and cyber security and data governance review tends to be treated as a post-implementation task. By the time those questions get asked, the architecture decisions are already made. The vendor is already processing patient data. The integration with the EPR is already live.

The same research makes clear this is accelerating. AI investment has moved from exploratory to operational from productivity tools, automation to decision support embedded in clinical workflows.

That shift means deeper integration with clinical systems, more patient data flowing through AI pipelines, and more third-party vendor access to NHS infrastructure. The security implications of all of that are rarely on the table when deployment decisions get made.

Data sharing across ICS boundaries is another pressure point. The 10 Year Plan and the neighbourhood working model both depend on data moving freely across Trusts, primary care networks, ICBs, and community providers. The access governance for shared data often is defined either informally or not at all.

And then there is the financial dimension. When a Trust is trying to close a deficit, the cyber security programme is rarely ringfenced. This is not unique aspect to any one organisation, rather it is a system-wide pattern, which effectively  means that the gap between AI adoption and security governance is not self-correcting. It compounds.

The SWOT

Here is our honest read of where things stand. It is  neither absolute,  nor a criticism of any individual Trust, it’s the pattern we see across the sector, and it’s one worth being direct about

Questions to ask

1. Do we have a defined process for reviewing the cyber security posture of AI tools before they are deployed? If not, who is accountable for that gap?
2. When data moves across ICS partners and community providers under our neighbourhood working model, who owns the access governance for that data?
3. If a cyber security incident affected your AI systems tomorrow, what would our clinical impact be, and how quickly could we isolate and recover?

The data is clear that NHS investment in both AI and cyber security is going up. But investment without governance sequencing does not close the gap, it just makes the gap more expensive.