Skip to content

AI Compass / News

UK: AI medical devices to be monitored across their whole lifetime

RA & medtech

The UK is moving AI medical devices from one-off assessment to lifecycle monitoring, and manufacturers selling there can start preparing now.

Source: Government backs recommendations of NHS doctors-led AI Commission (MHRA, GOV.UK),

Note: I provide engineering services in regulatory affairs for medical devices.

Technical drawing: on the left a one-time checkpoint before market entry, on the right a performance curve with measurement points that sinks towards a limit line and rises again after updates V2 and V3, below it a PCCP bracket spanning both updates.

What happened: On 6 October 2026, the UK government committed to implementing all 44 recommendations of the National Commission into the Regulation of AI in Healthcare. The Medicines and Healthcare products Regulatory Agency (MHRA) set up the independent, NHS doctor-led commission in 2025. One of its central conclusions: AI medical devices should be assessed and monitored throughout their working life, not mainly once before market access. By December 2026, the MHRA intends to publish draft guidance on Predetermined Change Control Plans (PCCPs), plans that define future changes in advance, for public comment. An implementation roadmap will follow by spring 2027.

What this means: According to the government response, post-market surveillance (PMS, monitoring the device once it is in use) is to be strengthened, including how manufacturers respond to performance changes in real-world settings. The PCCP draft is meant to let manufacturers set boundaries for allowable change instead of only pre-specifying individual changes, as long as the intended use stays the same. For CE-marked devices: Great Britain will keep recognising the CE mark indefinitely, but its stricter post-market surveillance rules also apply to CE-marked devices placed on the GB market since 16 June 2025.

What you can do:

  1. Put metrics into the PMS plan: Define which performance values of your AI model you measure in use and at which threshold you act. Drift, meaning declining performance because data or workflows change in use, belongs in the plan as an item of its own. The commission recommends escalation processes when performance degrades, even if no reportable incident has occurred.
  2. Describe changes up front: Document now which model updates are allowed, how you test them and how you roll them back. Your PCCP will later build on this.
  3. Make versions traceable: The MHRA plans guidance on unique device identifiers (UDI) and on version control across updates. Record which model version runs where.
  4. Consider the AI Airlock: Phase 3 of the MHRA’s regulatory sandbox focuses on post-market surveillance. Applications received before 31 October go into the first selection in November 2026, later ones into spring 2027.

Context: The new AI rules are not in force yet. The government plans to share details on post-market surveillance, reporting and version control by spring 2027.

What does this mean for your team? Let’s talk it through. Request an intro call