AI in the public sector should be straightforward: less admin, faster decisions, services that actually run. Yet in the UK, one name keeps turning every procurement into a culture war: Palantir.
On paper, the pitch is compelling—connect fragmented datasets, help hospitals manage capacity, reduce documentation time, coordinate social services. In reality, you get a volatile mix: sensitive citizen data, heavily redacted contracts, a US firm with deep defense/intelligence roots, and rising pushback from doctors, NGOs, and local campaigners.
Let’s cut through the noise: which contracts, what the recent numbers say, which criticisms are technically grounded, and how public bodies (or vendors) can deploy AI without getting torched.
What Palantir is—and why it triggers backlash
Palantir sells data platforms (Foundry, Gotham) designed to ingest, model, govern, and analyze data across dozens of systems. Translation: where an organization has 40 databases that don’t talk to each other, Palantir provides a layer to build dashboards, workflows, and models (including AI).
The controversy isn’t just “software.” Palantir’s brand is tightly linked to security services and defense. Even if the tool is neutral, public perception changes drastically when you apply it to:
- patient data (NHS)
- children and family case files (social services)
- defense and national security
That’s why the debate escalates fast—and not always rationally. But there are hard, factual issues too.
The deals at the center of the storm (with recent figures)
1) NHS England: the Federated Data Platform (FDP)
The biggest contract is NHS England’s £330 million deal over 7 years for a Federated Data Platform (FDP) intended to make operational and clinical data usable across the system (capacity, patient pathways, bed management, etc.). Source: Digital Health (contract published but heavily redacted) [digitalhealth.net, 2024].
Why it’s sensitive: the NHS has a long history of public distrust around data programs. Add:
- a US vendor
- a contract with key sections blacked out
- persistent fears of re-identification even when data is pseudonymized
The British Medical Association (BMA) warned it could undermine public trust in NHS data systems (source: The Guardian, July 2025).
Palantir’s UK leadership pushed back, arguing the FDP improves patient outcomes and that monetizing medical data would be illegal (Louis Mosley comments reported by The Guardian, July 2025).
2) Coventry City Council: AI in social services
In 2025, Coventry City Council signed an approximately £500,000 per year contract with Palantir to pilot AI in social services—transcription, case summarization, and support for SEND (Special Educational Needs and Disabilities) workflows. Sources: The Guardian (Aug 2025) and Computing.co.uk.
This touches extremely sensitive data: children, families, vulnerability indicators. Even as a 12‑month proof of concept, it triggered protests about ethics, governance, and Palantir’s defense links.
3) Defense / MoD: expanding footprint
Beyond health and local government, Palantir’s ties with the UK Ministry of Defence have been deepening. Reporting and sector analysis suggest Palantir’s UK public-sector contracts exceed £500 million, with potential commitments that could rise further.
Even if defense analytics is legitimate, it spills over into the NHS and council debates: for many critics, it’s “the same company” touching health, social care, and military work.
Why the controversy keeps happening: 6 concrete drivers
1) The real battleground is trust, not technology
You can have the best architecture in the world—if citizens believe you’re building a surveillance machine, you lose.
“Federated” is meant to sound safer, but the public hears: more data + powerful vendor = more risk.
2) Redacted contracts create a vacuum
If you publish a contract but black out large portions, you create a narrative vacuum—and opponents fill it with worst-case assumptions.
Legal and civil society groups have challenged transparency around these procurements (reported by The Register, 2024).
Even when redactions are defensible (security, trade secrets), the political effect is brutal.
3) Legal and compliance risk is messy in practice
The UK has strong data protections (UK GDPR, Data Protection Act). But “legal on paper” isn’t enough. People want clarity on:
- accountability when something goes wrong
- scope creep and secondary uses
- opt-out handling
- subcontractors and data flows
4) The “US company” issue: CLOUD Act and sovereignty fears
Even with UK hosting, a US company can face US legal obligations. A Swiss report in 2025 reignited debate about whether US intelligence services could access data held by US vendors, prompting UK MPs to call for more transparent due diligence (source: The Guardian, Dec 2025).
This is often technically misunderstood—but politically potent: “our health data with a US firm.”
5) Vendor lock-in and long-term cost
Palantir is known for shipping fast—and for becoming deeply embedded. Once your data model, pipelines, and workflows depend on a proprietary platform, switching becomes expensive.
Public sector buyers fear a familiar pattern: win short-term gains, pay for a decade.
6) Reputation baggage: surveillance and human-rights claims
Even if a local project has nothing to do with defense, Palantir’s brand carries “surveillance tech” associations. Amnesty International criticized the NHS award, citing links to serious human-rights concerns (source: The Guardian, Nov 2023).
You can dispute the framing, but you can’t ignore the impact: social license becomes as important as performance.
Why the UK still signs: the value is real
Let’s be honest: these deals exist because the pain is real.
- The NHS is fragmented across legacy systems and silos.
- Councils run overstretched teams with manual, repetitive workflows.
A well-implemented data platform can:
- reduce reporting overhead
- improve capacity planning
- accelerate operational decisions
- standardize governance
Budget pressure amplifies the push. The Financial Times reported a rapid increase in UK government AI-related contracts: £573m by 15 July 2025, already above £468m for all of 2024, tied to modernization and large projected savings (source: FT, 2025).
So yes: the state wants outcomes. And Palantir often executes where overpriced consultancies deliver slide decks.
Deepthix take: the core mistake is treating this as “just a tool”
Deploying data platforms in health or social care isn’t normal IT procurement. It changes the relationship between citizens and institutions.
If you don’t deliver:
- clear governance
- enforceable guarantees (contractual + technical)
- radical transparency where possible
…you hand critics an easy win. And sometimes they’ll be right.
How to deploy AI in sensitive public services without getting wrecked (practical checklist)
1) Separate “data platform” from “genAI” use cases
Don’t bundle everything. A federated data platform isn’t the same as an LLM summarizing case notes.
Action: separate DPIAs, separate governance, separate audit plans.
2) Publish what you can—especially the guardrails
If 60% is redacted, you’ve already lost.
Action: publish at minimum:
- data categories
- purposes
- retention periods
- subcontractors
- anti-monetization clauses
- audit rights and an exit plan
3) Build for minimization: less data, better controls
The safest data is the data you never collect.
Action:
- pseudonymize by default
- separate identifiers
- immutable audit logs
- role-based access + justification
4) Contract for reversibility (anti lock-in)
Action:
- open export formats
- documented data models
- annual migration drills
- capped, priced exit clauses
5) Measure outcomes—kill “eternal pilots”
Action: simple KPIs:
- case processing time
- discharge delays
- data entry error rates
- staff satisfaction
No measurement, no scale.
6) Bring frontline and patients in early
Basic, but constantly ignored.
Action: co-design sessions, veto rights on certain uses, and a clear incident/reporting channel.
Conclusion: the Palantir fight is a symptom
Palantir attracts controversy because it embodies what people fear: powerful analytics, defense history, perceived opacity. But the real issue is the public sector’s maturity in data governance.
If the UK wants more efficient services, it will need technology. The question isn’t “AI or no AI.” It’s: which guarantees, what transparency, what reversibility.
And if you’re a founder or vendor: in sensitive domains, technical execution isn’t enough. You must win the trust battle.
Want to automate your operations with AI? Book a 15-min call to discuss.