In late May 2026, the Mayor of London’s Office for Policing and Crime (MOPAC) blocked the Metropolitan Police’s plan to buy artificial intelligence software from Palantir — a contract worth roughly £50 million across two years (£25.3m in 2026-27 with an optional £24.8m extension). The Met called the decision “disappointing.” Palantir told MOPAC it intends to challenge the decision in court. Within days, the force published some thirty tender notices covering more than £300 million of future technology requirements, a signal that it intends to acquire similar capabilities by other means.

It is tempting to frame this as London “banning Palantir from frontline policing.” It is not that — and the gap between what happened and what the headline implies is worth sitting with, because the truth is both narrower and, in some respects, more troubling.

What was actually decided

The Met’s deal was not blocked on civil-liberties grounds. According to MOPAC, it was blocked on procurement grounds. The force had engaged with only one potential supplier — Palantir — and could not evidence a “reasonable or adequate” process of market testing to show this was the most economical option. MOPAC concluded the Met had not “adequately ensured, or demonstrated, value for money,” and that the force had failed to bring its procurement strategy to the Deputy Mayor for approval as required. The Mayor’s office stressed that with budgets this tight, “robust processes” matter when awarding contracts worth £25m a year.

So this was, on its face, a governance and value-for-money intervention, not a ban. Nothing in the decision says Palantir’s software is too dangerous for policing. The Met remains free to come back with a properly competed tender — and the £300m shopping list suggests it intends to keep the capability in view. A purist could call this a speed bump.

But that reading is too tidy. A single-supplier procurement that engaged with no one but Palantir, kept below the threshold that would have triggered mayoral oversight, is exactly the pattern that should worry anyone who cares about democratic control of policing technology. The fact that it took a procurement technicality — not a substantive debate about surveillance — to halt the deal tells you how thin the oversight is.

The deal that didn’t need permission

The blocked £50m contract was not the Met’s first dealing with Palantir. Earlier in 2026 the force ran a pilot using Palantir software to surface officer misconduct and corruption — fusing data from work phones and laptops, geolocation, shift-rostering logs and internal databases. The reported pilot value was £489,999: a figure that sits deliberately, conspicuously, just below the £500,000 line at which mayoral oversight would kick in. When the pilot was extended, the Met declined to disclose the cost, citing commercial sensitivity.

The capability is genuinely double-edged. The system reportedly contributed to three arrests for offences including sexual assault and fraud, with dozens more officers facing misconduct assessment — and few would mourn better tools for rooting out predators in uniform. But the Metropolitan Police Federation called the continuous geolocation tracking “highly intrusive,” warning it risks monitoring officers off duty, and described the project as “not proportionate, just or proper.” The Public and Commercial Services Union said it received no formal consultation. If a data-fusion platform can be turned inward on the police themselves with this little ceremony, the question of what it does when turned outward — on the public — is not academic.

Why the company matters

Civil-liberties groups have not been quiet, even if “ban” overstates their success. Liberty has warned of “dystopian predictive policing” and “indiscriminate mass surveillance” arising from police adoption of Palantir’s software. Reporting indicates Palantir has built real-time data-sharing networks across UK forces that pull in sensitive details of vulnerable victims, children and witnesses — not just suspects. And Big Brother Watch has lodged a complaint with the Information Commissioner’s Office after the National Police Chiefs’ Council instructed forces to “neither confirm nor deny” holding documents about Palantir in response to Freedom of Information requests. As Big Brother Watch’s Jake Hurfurt put it, the central FOI unit “is a danger to transparency and the public’s right to know.”

That secrecy is the through-line. The mechanism that almost let the £50m deal through without scrutiny, the threshold-skirting pilot, the “neither confirm nor deny” posture on FOI — these are not separate stories. They are the same instinct to build surveillance infrastructure below the waterline of public debate.

It also matters who Palantir is. The same firm is the subject of a parallel revolt over the NHS’s £330m Federated Data Platform, where in June 2026 a cross-party Commons committee urged the government to use a February 2027 break clause to exit the contract, citing over-dependence on a single foreign vendor. Internationally, the company is closely associated with US military and intelligence work and — as the ACLU and others have documented — with immigration enforcement and deportation operations. A data-fusion platform is not neutral plumbing; it embeds the worldview and the institutional clients of the company that builds it. Function creep is not a hypothetical risk with this kind of architecture. It is the business model.

The real lesson

So: not a ban. A blocked procurement, a court challenge looming, and a £300m list that suggests the underlying appetite is undimmed. Anyone claiming London has shut Palantir out of frontline policing is reading too much into a value-for-money veto.

But the episode is still a small victory for oversight, and a large warning. It shows that the controls exist — and that the Met repeatedly structured its purchases to slip beneath them. The right response is not to celebrate a ban that didn’t happen, but to insist that decisions about predictive policing, biometric data fusion and 24/7 location tracking are made in the open, on the record, with the public’s consent — and not settled in a procurement notice nobody was meant to read.