Khan Blocked a £50m Palantir Police Deal. The Settlement Did Not Award It Back—and the Savings Are Still Unproven.

Factrail investigation | London, United Kingdom | Decision: 21 May 2026 | Settlement: 14 October 2026 | Evidence checked: 31 October 2026

The finding. London’s police leadership sought a contract worth up to approximately £50 million for Palantir’s Unified Operational Analytics software. The deputy mayor responsible for policing refused approval on 21 May because, in her written assessment, the Metropolitan Police had not obtained the required sign-off for its procurement strategy, had seriously considered only one supplier and had not demonstrated value for money. Palantir challenged the refusal. On 14 October the company and the Mayor’s Office for Policing and Crime (MOPAC) withdrew their High Court claims, explicitly clarified that the decision was not based on Palantir’s values or ethics, and agreed the company could compete in a new procurement. No court decided that the original selection was unlawful; no court decided the refusal was unlawful; the original £50 million award was not reinstated. The unresolved public question is whether a new competitive process can substantiate the promised policing benefits and a credible price rather than merely reproduce the first choice. Primary refusal letter (opens in a new tab) | 14 October settlement statement (opens in a new tab).

The document that explains the veto

The most revealing record is not a press quote about Silicon Valley politics. It is the deputy mayor’s three-page correspondence released through the London Assembly. Kaya Comer-Schwartz’s 21 May letter to Metropolitan Police Commissioner Mark Rowley identifies an initial term of one year plus a possible further year, with a total proposed value of roughly £50 million. It says the force was required by the applicable Scheme of Delegation to obtain MOPAC approval for the procurement strategy before proceeding. The deputy mayor said that did not happen, notwithstanding an earlier reminder. That is an attributed institutional assessment, not a judicial finding of a statutory offence.

Her criticism was unusually specific. The Met had seriously considered or engaged with only Palantir; in her view it had not adequately tested whether another supplier could deliver the required services more economically. She also identified a preceding Palantir contract directly awarded and extended without an open competition, with a stated total value marginally below the threshold for MOPAC consideration. The letter does not establish that the earlier contract was unlawfully split to evade oversight, nor disclose the exact threshold or price. It does identify a structural risk: pilot contracts can become the technical and commercial foundation for much larger procurements before alternatives receive equivalent evaluation.

The letter provides a more precise financial trail than the headline £50 million. MOPAC had initially been told that the prospective arrangement could cost approximately £15 million to £25 million annually. After the Met’s direct negotiations with Palantir, the proposal landed at the upper end, around £25 million a year. The deputy mayor said the evidence presented did not satisfy her that both years could be financed without unacceptable pressure on other budget lines. The £50 million was therefore a potential two-year contract value, not a confirmed public payment to Palantir, an independently verified saving or a realized loss. MOPAC letter, pages 2–3 (opens in a new tab).

The chronology: a pilot, an attempted expansion, a court fight, a reset

The documentary chain begins before the dispute. Comer-Schwartz told the Assembly she had raised questions with the Met in September 2025 about proposed suppliers for precision policing, vetting, cybersecurity and human-in-the-loop safeguards. She was later updated on a time-limited Palantir pilot focused on culture and standards. The pilot fell below the financial threshold requiring her formal approval, according to her account. The proposed Unified Operational Analytics (UOA) arrangement would make some of that work permanent and expand it into operational policing. These are not interchangeable contracts: an employee-conduct pilot, an analytical policing platform and a future competitive UOA award have different scopes and approval questions.

On 21 May 2026 the deputy mayor refused approval for the proposed Palantir award. The London Assembly announced a committee examination on 16 June, citing the absence of meaningful competition and the value-for-money issue. In June Palantir began High Court proceedings. Its argument, as reported from court papers, was that MOPAC had imposed an unlawful veto and relied on an irrelevant political or ethical objection. MOPAC’s response was that it had exercised legitimate procurement and financial oversight. The Metropolitan Police defended its approach as legally and commercially compliant. These are competing legal and institutional positions, not findings that the courts adopted.

The public political record was awkward. An earlier mayoral spokesperson had discussed whether companies receiving public funds should share London’s values. Palantir said the remarks gave the impression that the firm’s politics had affected the procurement decision. The issue mattered because public authorities cannot simply invent a supplier-exclusion ground outside the applicable procurement framework. In a September answer to Assembly Member Gareth Roberts, the mayor explicitly recognized that the Procurement Act 2023 limits the grounds on which a supplier can be excluded.

On 14 October, before the planned trial, MOPAC and Palantir jointly announced that both had withdrawn the claims made in their pleadings. MOPAC stated that the May decision had not been based on the company’s ethics, values or news coverage and recognized that prior communications could have created a different impression. The joint statement also said there had never been a MOPAC policy against contracting with Palantir. The Met and MOPAC were continuing a fresh procurement, and Palantir was welcome to compete. Reuters and the Guardian reported that no damages were paid and each side would bear its own legal costs. None of those terms awarded Palantir the blocked contract. City Hall statement (opens in a new tab); Reuters, 14 October (opens in a new tab).

Who could decide—and who could stop the deal

The Met is the operational customer. Commissioner Rowley and his management were responsible for identifying a system to improve intelligence analysis, evidence processing and administrative productivity, then building a procurement and business case. They could propose the technical requirements and negotiate with suppliers. That does not mean they had unfettered authority to bind London to a £50 million award.

MOPAC is the governance and spending gate. Comer-Schwartz’s signed letter says the Scheme of Delegation required its approval of the procurement strategy and that this was not obtained. She could withhold approval for the proposed award, as she did. Her authority should not be conflated with a personal statutory power of the mayor to blacklist a technology company. Sadiq Khan was publicly accountable for the policing office and its policy direction; the record identifies the deputy mayor as the signatory of the actual refusal. London Assembly members—including Susan Hall, who pressed the mayor about operational consequences, and other members questioning supplier ethics—could scrutinize, demand documents and debate the decision, but not themselves sign the commercial award.

Palantir was the proposed supplier and later a claimant. It could offer software, make commercial representations, litigate the refusal and compete again. It could not unilaterally override the public authority’s approval process. The High Court was a potential legality checkpoint, not a procurement committee: the October settlement ended the dispute before a merits ruling. The next actual veto points are the design of a defensible tender, any supplier exclusions supported by law, assessment of competing bids, affordability approval, data-protection controls and the final award decision.

The legal distinction is important. Government guidance on the Procurement Act describes specific mandatory and discretionary grounds for excluding suppliers, and requires fair treatment and competition. It does not say that public authorities must buy from every controversial supplier. Equally, disapproval of a company’s leadership or other government contracts is not automatically a lawful ground for exclusion. MOPAC’s October correction narrows the reported basis of its May decision; it does not erase the documented procurement concerns. Cabinet Office exclusion guidance (opens in a new tab); London Assembly answer, September (opens in a new tab).

The money: a contract ceiling, an opportunity and an untested counterfactual

Who stood to receive money is clear at the first level: Palantir would have been the vendor under the proposed UOA award. The disclosed price, however, is not proof of revenue recognized, profit earned or public cash disbursed. Nor is the amount an independently established market price. The refusal letter’s £15 million–£25 million annual range and £25 million negotiated endpoint show why comparative bids and cost breakdowns mattered. A tender might produce a lower price, a different architecture, a different risk allocation or the same supplier on defensible terms. Until a completed process is disclosed, those are possibilities, not results.

The potential public beneficiaries were investigators, analysts and London residents if the system shortened evidence searches and enabled better crime investigation. The potential public payers were London’s policing budget and, indirectly, taxpayers and service users if expenditure displaced other staffing or operational needs. A vendor that already supplies a pilot has advantages: familiarity with data, workflow integration, training and incumbent experience. Those advantages can reflect genuine performance, but also create switching costs. They do not prove unlawful favoritism. A credible procurement should distinguish a demonstrably unique capability from a requirement written around an incumbent’s existing product.

Financial stakes also extend beyond this single award. Palantir sells analytical software across government, and the company has faced separate public scrutiny over contracts in health and national security. Those other arrangements are context for its commercial incentive, not evidence that the Met’s specific £50 million would have funded a particular overseas military or immigration operation. The hiring of former Labour deputy leader Tom Watson as a lobbyist, reported before the settlement, illustrates political access and public-relations strategy; it does not prove he determined the legal outcome or received a fee tied to this contract. There is no documentary basis here for alleging an illicit payment to Khan, Comer-Schwartz, Rowley or any official.

The Met's strongest case: a safety and productivity problem, not merely vendor preference

The Metropolitan Police did not describe UOA as an optional branding exercise. Its leaders said the software could automate labor-intensive intelligence analysis, speed the processing of mobile-phone material and free personnel for frontline tasks while budgets were tight. Commissioner Rowley argued the original procurement approach was legally and commercially sound. Reporting on the force’s business case described possible savings of approximately £75.5 million over two years, but this remains a management forecast in the sources reviewed, not a measured net saving. Any fair analysis must compare the forecast with the full purchase, implementation, security, training and maintenance costs and with what other suppliers could deliver.

The policing pressures were real even if the technology forecast was contestable. Reporting from the London Policing Board described a 2026–27 plan involving reductions of around 1,150 jobs, with work equivalent to roughly 500 positions expected to be addressed through AI tools. That does not mean precisely 500 frontline officers would be saved by a Palantir contract, or that blocking it caused all projected losses. A budget plan, a staffing scenario and an independently demonstrated causal impact are three different things. A May decision cannot retroactively explain every financial pressure established in earlier budget discussions.

At a June Assembly exchange, Susan Hall challenged Khan about whether the pilot could continue and whether stopping the contract jeopardized frontline policing. Khan replied that difficult choices had already been expected in 2026–27 and 2027–28, and argued that failed IT procurement can waste funds needed for officers. That exchange supplies the strongest competing public cases in the protagonists’ own words: urgency and crime-fighting capability on one side; market testing and control of long-term spending on the other. The answer is not established by which politician sounds more alarmed. London Assembly transcript (opens in a new tab); Independent, 14 October (opens in a new tab).

The pilot trap and the operational data question

The earlier culture-and-standards pilot is relevant because it shows why a small technology contract can have a large downstream effect. The deputy mayor’s May correspondence says she did not approve that pilot because it fell below her formal threshold. It also says a directly awarded and extended Palantir arrangement sat just below the level requiring MOPAC consideration. A journalist can legitimately ask whether approval thresholds, extension rights and system design created a de facto incumbent before a larger tender. But the present documents do not establish a deliberate avoidance scheme, and no regulator or court has made that finding.

The pilot’s reported use of data about approximately 45,000 police personnel to identify potential misconduct or welfare risks also raises a distinct accountability issue. Claims that the software flagged cases are not equivalent to proof that each flag was accurate, fair, independently reviewed or useful. The number of records ingested is not a number of people found guilty. Staff privacy, lawful data access, audit logs, false positives and human review all require a separate documented assessment. The deputy mayor’s earlier questions about cyber security and human-in-the-loop safeguards suggest those were known concerns before the public contract dispute.

An open competition should test the whole system, not just the vendor’s headline algorithm. Which source datasets would be joined? Who can search them? What decisions remain human? How will investigators explain an output to a court or a defendant? What happens to police records when a contract ends? Could the authority migrate to a new supplier without paying for bespoke extraction? These questions matter because a low initial licence price can become expensive if the authority loses the practical ability to leave. No current source establishes the final answers for a future UOA procurement. Assembly release and correspondence (opens in a new tab); Guardian pilot reporting (opens in a new tab).

The ethics controversy: politically relevant, legally different

Palantir’s US government work and public positions associated with its leadership were part of the political controversy surrounding the London contract. Those concerns are legitimate subjects for democratic debate about public technology, civil liberties and the use of taxpayer money. But a political objection and a legally applicable exclusion ground are not synonyms. The May press discussion of London’s values was precisely what Palantir said had made the refusal appear discriminatory. In October MOPAC explicitly corrected that impression and affirmed the company’s eligibility to compete.

This does not require readers to accept Palantir’s every account of the procurement or MOPAC’s every earlier assertion. Palantir’s legal claim had asserted an unlawful veto. MOPAC had accused the Met’s proposed selection process of failing its procedural and financial tests. Both sides withdrew those claims without adjudication. Neither set of pleaded claims became a judgment. The strongest evidence for what MOPAC actually communicated at the time remains the signed 21 May letter; the strongest evidence for the agreed public position after settlement is the 14 October joint statement. A report that quotes only one of those records would obscure the change in public framing.

Khan’s October statement praised Palantir’s London jobs and investment, a marked contrast in tone with earlier debate over values. The contrast is politically meaningful, but motive remains unproven. It could reflect a negotiated correction of public language, a legal-risk calculation, a desire to preserve competition or several considerations. No released document shows a secret promise of a future award, a quid pro quo or improper influence. The only verified forward commitment is equal treatment in the renewed procurement. MOPAC settlement (opens in a new tab); Reuters (opens in a new tab).

What a serious value-for-money test would disclose

The central weakness in the public record is not that no official has offered a rationale. Both sides have. It is that the documents needed to compare them are incomplete. A defensible assessment would publish, with necessary security redactions, the original requirements and evaluation criteria, the 2025 pilot’s scope and full expenditure, its renewal decision, the UOA business case, market-engagement records, independent price benchmarks, the data-protection assessment and the new tender’s scoring methodology. It would separate expected analyst-hours saved from actually verified productivity, and estimated staffing substitution from realized reductions in overtime, vacancies or operational delay.

The procurement should also account for transition costs. If Palantir is the best technical option, a fair process can demonstrate that rather than assume it. If another supplier can meet the same policing need, the force should quantify the integration and migration costs honestly instead of hiding them as reasons to preserve an incumbent. Security-sensitive information may legitimately remain confidential; secrecy over sensitive tactics does not require concealing every spending assumption or governance decision. The Assembly’s scrutiny is particularly important because the operational customer and spending overseer disagreed so sharply.

The outcome indicators should be concrete: cost per completed evidence-analysis task; median time to process devices and intelligence requests; backlog of investigations; validated false-positive rates in misconduct screening; actual software expenditure and extensions; competitive bids received; and the number of frontline posts or staff-hours preserved after adjusting for other budget decisions. Without these, both the £75.5 million savings narrative and the argument that the veto prevented waste remain counterfactuals. They may be plausible; neither has been independently established by the records examined.

The strongest counter-case to this investigation

A reasonable critic of the veto can say the Met faced immediate public-safety and workforce pressures, had tested Palantir’s software in real workflows, and considered an unusually specialized tool for which alternatives might not be equally capable. Prolonging procurement may itself carry costs. The Met said its process complied with the law. Palantir could reasonably object if a public authority appeared to punish it for the political opinions of executives rather than applying published procurement criteria. The settlement’s clarification about ethics is a substantive concession to that concern.

A reasonable defender of the veto can answer that a successful pilot does not excuse missing required approval or inadequate competition on a much larger commitment. The deputy mayor’s contemporaneous letter identified concrete process and affordability defects, not only public-relations discomfort. £25 million a year is a material commitment in a constrained policing budget. The risk of costly dependence is precisely why the value-for-money test belongs before, not after, an award. The public record supports taking both arguments seriously. It does not support a categorical claim that Palantir was the only possible vendor or that MOPAC acted illegally.

What remains unknown, and who must answer

As of 31 October 2026, the documents reviewed do not establish a final UOA supplier, a signed replacement contract, a competitive final price, a court judgment on the withdrawn case, verified net savings, or a full account of payments under the earlier pilot. A reported spring 2027 decision window is a forecast, not a completed award. The most consequential missing documents are the original Met procurement strategy, competing technical options, the signed pilot and extension terms, detailed financial model, any independent data-protection and equality assessments, and the final tender results.

The accountability map is therefore still open. Rowley and the Met must substantiate operational need and technical performance. MOPAC and Comer-Schwartz must show that the new process tests competition, affordability and lawful supplier treatment. Khan must answer for political communications that complicated the dispute while maintaining clear separation between values rhetoric and procurement decisions. The Assembly can demand evidence and expose gaps. Palantir and rival bidders must compete on measurable terms. Londoners must ultimately judge the outcome through audited costs, rights protections and actual policing performance—not a headline contract ceiling or an untested claim that AI automatically saves officers.

Bottom line: The October settlement resolved a legal confrontation, not the underlying procurement economics. The mayoral office kept its refusal in place, Palantir kept the right to compete, and the public still lacks the independent figures needed to know whether the first £50 million proposal would have delivered value. The next award, not the settlement announcement, will show whether London’s institutions can convert a politically charged technology fight into verifiable public accountability.

Related Factrail context

The public-finance test behind this policing purchase is part of the wider UK fiscal sustainability discussion (opens in a new tab): a large proposed contract can reduce room for other public services unless officials show a defensible net benefit. The Institute for Fiscal Studies (opens in a new tab) provides a useful institutional context for distinguishing budget allocations from realized savings; neither its work nor the MOPAC record proves that this particular software purchase would save £50 million. London's procurement trade-offs also connect to cost-of-living pressure (opens in a new tab), since taxpayers and service users ultimately bear the opportunity costs of public spending. These links provide analytical context rather than claims that those institutions decided the tender.

Sources and documentary basis

1. Kaya Comer-Schwartz, 21 May 2026 correspondence and refusal letter, released to the London Assembly (opens in a new tab) — primary record of the veto, approval rule, market-testing critique and price.
2. MOPAC and Palantir, joint statement, 14 October 2026 (opens in a new tab) — primary settlement terms and corrected explanation.
3. London Assembly Police and Crime Committee, 16 June 2026 notice (opens in a new tab) — oversight scrutiny.
4. London Assembly, mayoral answer on Palantir supplier exclusions, 10 September 2026 (opens in a new tab) — mayor's stated legal constraint.
5. London Assembly, mayoral answer on a revised UOA procurement, 10 September 2026 (opens in a new tab) — procurement process response.
6. London Assembly, oral exchange with Susan Hall and Sadiq Khan, 18 June 2026 (opens in a new tab) — competing operational and governance arguments.
7. Cabinet Office, Procurement Act 2023 guidance on exclusions (opens in a new tab) — applicable statutory framework guidance.
8. Cabinet Office, competitive tendering procedures guidance (opens in a new tab) — procurement principles.
9. Reuters, Palantir and London mayor's office settle lawsuit, 14 October 2026 (opens in a new tab) — independent account of settlement.
10. Financial Times, Sadiq Khan accuses Met of trying to fix contract, July 2026 (opens in a new tab) — court pleadings and competing positions, claims not adjudicated.
11. Financial Times, Palantir drops legal case, 14 October 2026 (opens in a new tab) — settlement and pilot context.
12. The Guardian, Met gets extension to AI project, 24 June 2026 (opens in a new tab) — pilot and personnel data.
13. The Independent, London mayor's office resolves Palantir dispute, 14 October 2026 (opens in a new tab) — staffing and budget positions.
14. The Guardian, Palantir and Sadiq Khan drop legal battle, 14 October 2026 (opens in a new tab) — political response, litigation settlement and lobbying context.

Editorial verification: primary MOPAC refusal and settlement checked against London City Hall records; court pleadings and signed pilot agreement were not independently obtained. This is an editorial draft, not a claim of publication.