Palantir NHS objections make patient consent an operating risk

0:05Host:You're listening to Leaders Insights. Today's subject: Palantir NHS objections make patient consent an operating risk. Forty-four thousand people filing paperwork to get their own health records out of one platform. Is that a protest or is that a data operations problem?

0:21Expert:Both, and the second one is the part my peers underestimated. More than 44,000 people have filed legal objections to stop the NHS using their health records on the Palantir-powered platform, coordinated by the not-for-profit 38 Degrees, invoking Article 21 of the UK GDPR, which is the right to object to your data being processed, and that was reported by Computing. Each of those is a request someone has to assess.

0:50Host:Define Article 21 for people who do not live in UK privacy law.

0:54Expert:It is the clause that says if you object to processing based on public task or legitimate interests, the organisation must stop processing your data unless it can show compelling legitimate grounds to continue. The Guardian reported that. It is not an off switch, it is a burden shift. The controller has to actually justify itself, case by case.

1:15Host:Case by case, forty-four thousand times. Nobody has that team.

1:19Expert:Nobody has that team. And the objectors also asked the NHS to pause processing their data while the requests are being weighed, per the Guardian's Robert Booth. That pause request is the operationally nasty bit, because the platform is live in clinical workflow. You cannot quietly hold forty-four thousand records in limbo inside a scheduling system.

1:40Host:What is the platform actually doing? I want to be precise before we judge it.

1:44Expert:The Federated Data Platform, FDP, connects information across the health service, things like surgery scheduling and discharge planning. Computing reported it was operational at 142 NHS trusts, with 137 reporting benefits. A trust is an NHS organisation running hospitals or mental health services. So this is not a research sandbox, it is in the operational path.

2:09Host:NHS England's answer?

2:11Expert:Short and legalistic. It said it would respond to the objections under data protection law, and that all NHS organisations remain in control of their data within the FDP, with suppliers unable to access it for their own purposes. That is factually about the contract. It is not an answer to the trust question, and I think they know that.

2:31Host:Let me be uncomfortable. Is this just anti-Palantir politics dressed up as data governance?

2:37Expert:Partly, yes, and I will not pretend otherwise. The campaign cites distrust over Palantir's work with the Israeli military and US immigration enforcement. That is a political objection. But the mechanism chosen is a legal one, and legal mechanisms do not care about your motive. Once someone files under Article 21, you owe them a reasoned response regardless of why they filed.

3:01Host:Can the NHS simply refuse all of them?

3:04Expert:Probably, on direct care grounds. That is my reading, not a fact. What is fact: health minister Preet Kaur Gill confirmed in June that individual patients in England cannot opt out of having their data processed by the FDP, while individual NHS trusts can opt out as institutions. She said the National Data Opt-Out, the standard NHS mechanism for refusing data sharing, does not currently apply to FDP products, because in most cases data is being used for direct care.

3:36Host:So patients have no exit, hospitals do. That asymmetry is going to age badly.

3:42Expert:It already is. At a scrutiny meeting reported on 2 October, medical professionals warned councillors that if patients lose trust they opt out, of data sharing and of care, which harms research, service planning and the safety of those individuals. That is the lesson for any data leader. When people cannot opt out of the system, they opt out of the service. Your dataset degrades in a way no access control fixes.

4:09Host:Now the numbers. Palantir says the thing works. Does it?

4:12Expert:Here is the claim, then the challenge. Computing reports Palantir saying the platform helped NHS trusts deliver 117,000 additional operations, cut long-stay patient discharge delays by 14.3% and improve 28-day cancer diagnosis rates by 5.6%.

4:31Host:Those are specific. Specific usually means defensible.

4:34Expert:Specific means measured, not causal. The Office for Statistics Regulation, the UK's statistics watchdog, urged NHS England to communicate the reported benefits more carefully, and NHS England added a caveat on 6 June acknowledging it could not establish a direct causal link because other variables had not been controlled for. Computing reported that. The organisation publishing the number conceded the number does not prove what it was used to prove.

5:03Host:That is an uncomfortable sentence for every dashboard I have ever signed off.

5:07Expert:It should be. And the distribution is worse than the headline. Freedom of Information data obtained by the campaign group Foxglove found 41 NHS trusts using the inpatient scheduling module, and 13 of them, about 30 percent, reported carrying out fewer operations overall since using the tool, according to The Register. On the outpatient side, Chelsea and Westminster accounted for 183,061 of the 217,846 patients removed from waiting lists.

5:42Host:One trust carrying the national story.

5:44Expert:One trust carrying most of it. That is a benefits-realisation failure, not a Palantir failure specifically. Any of us could publish an aggregate that hides a bimodal distribution. Most of us have.

5:56Host:Give me the strongest argument on the other side. Steelman it.

6:00Expert:The strongest counter is this. Staffing shortages, more complex cases, or pressure on hospital bed capacity might explain the fall at those trusts, as The Register noted. Confounders cut both ways. If you will not credit the platform for improvements because of uncontrolled variables, you cannot blame it for declines either. And ripping out a platform embedded in 142 trusts has a real patient cost that no one is putting a number on.

6:29Host:Who inside the NHS is saying what, data-side?

6:31Expert:This is the detail I would put in front of my board. The Chief Data and Analytics Officers Network, which represents senior NHS data staff, wrote urging the government to consider the break clause, saying the platform has been unable to demonstrate the level of evidence, confidence and realised value that should be expected following sustained investment and extensive policy support. That is reported via the Guardian. The practitioners, not the campaigners.

6:59Host:And the clock.

7:00Expert:A cross-party group of MPs told the government to prepare for life after Palantir, with the Health and Social Care Committee telling the Department of Health and Social Care to start preparing to replace the FDP when the contract reaches its break clause in February 2027, per The Register. Committee chair Layla Moran said the government's arguments for sticking with the platform had unravelled and it was time to prepare an alternative in time for spring 2027.

7:28Host:If you are a chief data officer in a bank or a retailer watching this, what transfers?

7:34Expert:Three things, and the first is the cheapest to fix. Your supplier's reputation is now part of your lawful basis in practice, because objections scale when a vendor becomes a political symbol. Second, benefits claims you publish will be audited by people with FOI rights and a motive. Third, consent and objection handling is a throughput problem, not a policy document.

7:57Host:Make that concrete. What do I do this week.

8:00Expert:Take your single most visible AI or data platform and answer one question in writing: if ten thousand people objected to processing tomorrow, what is the exact workflow, who signs the refusal, and how many days does it take. If you cannot name the owner and the service level, you have the same exposure NHS England has, at smaller scale. Then look at your last published benefits number and check whether the methodology note says before and after, or says controlled comparison. If it says before and after, add the caveat yourself before someone adds it for you.

8:37Host:This episode draws on Thousands demand NHS halt use of patient data on Palantir platform, 44,000 object to Palantir’s NHS data platform, the Guardian reports, MPs tell NHS to start packing Palantir's bags ahead of 2027 contract break, Pressure mounts from MPs on Palantir’s role in the NHS, NHS Palantir claims face scrutiny after data suggests uneven results, Patients Could Opt Out of Care Over NHS Data Platform Fears, Councillors Warned - This is the Coast. We'll stop there. The CDO tools for making the call are at mba-training.com.