Organisational attendees

• National Voices
• The NHS Alliance
• Patient Public Voices
• Understanding Patient Data
• NHS England
• Office of the National Data Guardian
• FDP Programme Team (NHS England)
• Information Commissioner’s Office
• British Medical Association
• Association of Medical Research Charities
• Nottingham University Hospitals
• NHS Confederation
• Chelsea and Westminster Hospital NHS Foundation Trust

Apologies received

• Royal College of Surgeons
• Royal College of General Practitioners
• East Kent Hospitals NHS Foundation Trust
• Patient Public Voices
• Healthwatch England

Open actions

Meeting minutes

Welcome and introductions

The Chair, Jacob Lant, opened the meeting. Apologies were noted.

The Chair noted that the Patients’ Association would be stepping down from the meeting and expressed thanks for the organisation’s valuable contributions to the group.

Minutes and actions from previous meeting

The group ratified and approved the minutes of the meeting held on Friday 20 February 2026.

Action log

The group reviewed the FDP action log and noted progress across several areas. It was confirmed that the DPIA for the AI Assisted Discharge Summary remained on track but could not be finalised until the product reaches pilot stage.

A more detailed update on FDP deployments, including the scale of deployments and usage across regions and organisation would be brought to the next meeting.

The NHS England FAQ has been updated to confirm that residency status data is not processed within FDP products, closing this action.

The group also noted the attendance of two representatives from Chelsea and Westminster Hospital to enable a more detailed discussion on the AI Assisted Discharge Summary product.

FDP and PET programme update

The Director of Data Management and Transformation provided an update on key activity to date across the Federated Data Platform (FDP) and the wider NHS England digital portfolio, outlining progress, current priorities and forthcoming milestones.

A member from Nottingham University Hospitals asked for an overview of the Solution Exchange (SolEx) on the FDP. It was noted that the SolEx operates as a marketplace model, enabling third-party suppliers to develop and offer solutions on FDP using shared technical capabilities. Suppliers self fund development, with local organisations procuring solutions directly, supported by NHS England guidelines and an evolving assurance framework.

The Chair asked if there were trending themes relating to slower uptake in FDP across some Trusts. The Director of Data Management and Transformation noted that variation was largely driven by differing local priorities and pressures, including competing programmes such as EPR implementations, as well as the presence of existing local solutions. It was acknowledged that these factors reflect locally determined needs and contexts rather than a single systemic issue.

The group noted that, alongside operational factors, some organisations may also hold concerns relating to the FDP supplier which could influence perceptions, trust and engagement, and this was recognised as a consideration within wider stakeholder sentiment.

The group received and noted the update.

FDP communications and engagement update

The FDP Senior Lead Communications and Engagement Manager provided an update on recent communications and engagement activity aimed at building public trust, responding to media interest, and strengthening transparency around the Federated Data Platform.

The Chair queried whether stakeholder engagement included the use of anonymous surveys to better understand how messaging around FDP was being received. It was confirmed that engagement with Trusts to gather feedback on their experiences of using FDP had been commissioned.

A member from the Association of Medical Research Charities inquired about the Westminster Hall debate on FDP, scheduled for Thursday 16 April 2026. In response, the Senior Comms Manager confirmed that ministers had been briefed, with a link the session shared in the meeting chat.

The group received and noted the update.

Single Patient Record (SPR) update

The SPR team joined the meeting and outlined the completion of a comprehensive technical options appraisal, noting the approval for a hybrid “emergent architecture” approach to enable iterative, co-productive development and adaptability. It was noted that initial service transformation areas, including maternity and frailty pathways, had progressed through early phases and are moving into alpha development, supported by continued clinician engagement and benefits validation. Commercial activity was underway, with supplier webinars, roundtables and engagement informing procurement specifications and identifying key challenges. Also highlighted was the comms and engagement plan, which remained aligned to the NHS modernisation agenda, incorporating stakeholder webinars, increased social media presence, to build trust and address misinformation, as well as cognisance of work being progressed outside the formal SPR programme, such as OneLondon’s independent proof of concept work.

The group noted that while the SPR update was helpful, it lacked sufficient detail to enable meaningful challenge. Members emphasised the need for earlier engagement and more substantive content to fully utilise the group as a trusted forum.

A member from the British Medical Association (BMA) raised concerns regarding Department of Health and Social Care (DHSC) social media communications during the recent Resident Doctor industrial action and highlighted the importance of maintaining public trust and confidence, noting the potential risks where communications may be perceived as inaccurate or lacking balance. In response, the Deputy Chief Data & Analytics Officer confirmed that work on data controllership within SPR remains ongoing, with no decisions taken, and is being progressed with the Government Legal Department and relevant teams. It was also reiterated that NHS communications are subject to established governance processes, guidance and senior review.

Action: Share the draft SPR comms engagement plan in the next meeting.

Action: Provide an overview of all engagement streams related to SPR, including the latest on who is involved, how to become involved, and terms of reference where available, to provide members with further engagement opportunities within the programme.

The group received and noted the update.

FDP Q3 benefits

The Q3 benefits update was not presented due to the absence of the presenter, and it was agreed that Q3 benefits would be deferred and incorporated into a consolidated annual benefits update at the next meeting.

AI-Assisted Discharge Summarisation (AADS) product update

Representatives from Chelsea and Westminster Hospital NHS Foundation Trust (ChelWest) and the Deputy Director of AI Delivery joined the meeting. The DD of AI Delivery opened with a statement welcoming the MHRA Announcement of additional funding for AI Airlock for the next 3 years, something the FDP is already involved with and welcomes ongoing debate around the future of AIaMD regulation.

An overview on the AI-Assisted Discharge Summarisation (AADS) product was provided. AADS is an AI-assisted, clinical administrative support tool that generates a first draft of the narrative section of the discharge summary. Clinicians remain fully responsible for the content of the discharge summary and use the AADS workflow to review, edit and finalise the content before the discharge summary letter is shared with the patient and their primary care provider. The group noted that the target benefits for AADS are to reduce the time clinicians spend writing discharge summaries, improve the consistency and quality of information provided to patients and primary care, and to support more timely discharge.

A member from the British Medical Association (BMA) challenged whether the current classification of AADS remained appropriate as AI medical device regulation continues to evolve. The FDP Programme team acknowledged that the regulatory landscape was evolving but provided assurance that AADS remained compliant with current legislation and regulatory advice. It was noted that the product is classified as a Class 1 medical device, with this position confirmed in January 2025 through NHS England Regulatory Affairs team and discussed with the Medicines and Healthcare products Regulatory Agency (MHRA), and that ongoing engagement with the MHRA and wider AI regulatory initiatives will be maintained to ensure continued compliance.

A member from the Office of the National Data Guardian raised a concern regarding human factors, specifically the risk that clinicians may, over time, become over reliant on AI-generated content or reduce scrutiny as familiarity with the tool increases. The FDP Programme team acknowledged this as a known risk and outlined planned mitigations, including ongoing monitoring, audit sampling over time, and the introduction of additional controls where review times or editing behaviours indicate potentially unsafe activity. The Chair emphasised the importance of maintaining clarity on the long term trajectory of the product, highlighting in particular the need for a clearly articulated plan for post market surveillance, alongside continued critical oversight in this area.

A member from the Public Patient Voices (PPVs) queried whether patients had reported feeling better informed and experienced smoother discharges. Whilst quantitative patient experience data was not yet available, the FDP Programme team reported early indications that patients were receiving more comprehensive and readable discharge summaries. It was also noted that anticipated benefits included faster completion of summaries closer to the point of discharge, which may support earlier discharge, although this was not yet evidenced. Members emphasised the importance of capturing direct patient feedback and robust evaluation, rather than relying solely on system or clinician reported benefits, to inform future scaling decisions.

A member from the Association of Medical Research Charities asked for clarification on data usage within AADS. It was confirmed that the product uses commercially available pre-trained models and does not train on patient data. It was further noted that all data processing takes place within a secure, UK based environment, with no retention of patient information by the models.

ACTION: Provide the group with a schematic and clear explanation of how patient data is processed within the AADS, including its interaction with LLMs.

The ChelWest team provided a demonstration of the training version of the AADS product, which uses synthetic data, showing how clinicians select patients, review clinical notes and AI- generated content, edit the draft text produced by the AI, and manually draft the sections that are out of scope for the AI (that is, narrative around medication changes, the summary of the verbal information provided to the patient, and safety netting).

The team also showed how the detailed clinical standards for what information should be in each section of the discharge summary can be easily accessed in the tool as well as the additional safeguards that are in place to support and facilitate thorough review of the draft text by the clinicians. Positive feedback was reported from clinical users, with no clinical safety issues identified to date (confirmed by thorough, post-discharge audits of each version of the discharge summary), and confirmation of significant quality improvements for discharge summaries drafted with AADS when compared to current practice (evidenced through a rigorous clinical performance study). The team also noted that there was ongoing refinement of the product informed by user feedback and post-discharge auditing activities.

A member from the British Medical Association (BMA) challenged the fundamental appropriateness of using generic large language models to support tasks traditionally undertaken by trained clinicians, expressing concern as to how such models could be considered suitable for producing clinical discharge documentation, even in an assisted capacity. In response, a member from ChelWest emphasised that the product has been clinically led from the point of project initiation, with clinical teams playing a key role in the design, testing and ongoing assurance of the product. The ChelWest team emphasised that AADS supports administrative drafting rather than replacing clinical decision making, with clinicians retaining full accountability through review and sign off. The tool was described as an evolution of existing discharge processes, with improved oversight and quality, and evidence of 30–40% time savings enabling more patient focused care as well as significant quality improvements. It was also noted that providing a secure NHS solution reduces risks associated with the use of unregulated external AI tools.

A member from the Information Commissioner’s Office (ICO) noted from an information governance perspective that the greater risk lies in clinicians using AI tools outside a governed NHS environment, rather than through a controlled solution such as AADS. The member therefore framed AADS as a positive, risk reducing intervention, emphasising that a thoughtfully designed, NHS governed solution was preferable to unregulated, ad hoc use. The member further noted the importance of considering the risks of not deploying such a solution when balancing safety, governance and assurance concerns, and expressed support for the programme’s approach in providing clinicians with a safe, approved pathway for AI-assisted tools within existing governance frameworks.

The Chair noted that the discussion reflected a range of perspectives within the group regarding the AI-Assisted Discharge Summarisation (AADS) product, highlighting the significance and sensitivity of the topic.

Any other business and close

The Chair noted an upcoming report from National Voices, Healthwatch England and the King’s Fund on NHS administrative challenges, highlighting its relevance to discussions on discharge processes and administrative burden.

The Chair advised the group that the next Data Transformation Check and Challenge meeting will be held on Friday 15 May 2026.