Palantir's fight with Sadiq Khan is no longer just a dispute over one policing contract. It is becoming a test of how far politics, procurement rules and public trust can shape the market for AI in government.
Palantir has moved toward legal action after Sadiq Khan blocked a proposed £50m Metropolitan Police contract, turning a local procurement row into a warning sign for every AI company trying to sell into sensitive public services.
The US data analytics group has sent a pre-action legal letter to the Mayor's Office for Policing and Crime, known as Mopac, challenging the decision to stop the Met from expanding its use of Palantir technology. According to The Times, the company argues the veto was unlawful and driven by subjective judgments about its values rather than the merits of the policing work.
That matters because this was not a small software upgrade. The Met wanted Palantir's systems to help process intelligence in criminal investigations, an area where police forces are under pressure to move faster through large volumes of digital evidence, phone records, emails and internal data. Scotland Yard has already defended the technology, saying it has helped identify misconduct and corruption inside the force in ways it could not easily replicate with existing tools.
Khan's office has framed the decision differently. The mayor did not simply object to Palantir as a company. Mopac said the Met had failed to follow the right process, had not secured the required approval for its procurement strategy, had only seriously engaged Palantir and had not shown enough evidence that the deal represented value for money. In public-sector contracting, those are not technicalities. They are the guardrails that stop a trial from becoming a dependency.
The difficult part for Palantir is that the procedural argument now sits alongside a broader political one. Khan has previously said Londoners expect public money to go to companies that reflect the values of the city. For Palantir, that is a dangerous standard because the company is already a lightning rod for criticism over its work with defence, intelligence, immigration and policing agencies.
Louis Mosley, Palantir's UK chief executive, has pushed back hard. In a Times Radio interview, he accused Khan of putting politics above public safety, arguing that the focus should be on protecting Londoners and helping police tackle serious threats. That line will appeal to police leaders who believe AI is now essential to modern investigations, especially when budgets are tight and criminal data trails are getting larger.
But the mayor's supporters see the same facts another way. If a police force can begin with a smaller contract, become familiar with one supplier and then seek a much larger deal without proper competition, critics will call that vendor lock-in. Palantir's opponents have repeatedly warned about a so-called land and expand model, where a company earns trust through limited deployments and later becomes hard to replace.
This is why the legal challenge is bigger than Palantir and the Met. AI companies selling to government are entering a market where winning the technical argument may not be enough. They also need to survive public scrutiny, political identity tests, data sovereignty concerns and procurement challenges from officials who are increasingly wary of being locked into powerful foreign platforms.
Govtech founders should price in political risk
For founders and investors, the clean lesson is that public-sector AI revenue is attractive, but it is not simple. A police, defence or health contract can create credibility that private customers respect. It can also put a company under a microscope, where old contracts, executive comments, geopolitical associations and campaign pressure become part of the sales process.
Palantir is not short of UK government work. It has a £330m NHS England contract for the Federated Data Platform and a £240m Ministry of Defence deal, while police forces outside London have also used its tools. That makes the London veto more striking. The British state is already using Palantir in major public systems, yet one of its most visible local leaders has decided this particular policing expansion should not go ahead in its current form.
The Met's position is also worth taking seriously. Police forces are dealing with huge digital workloads, from domestic abuse cases to organised crime investigations and internal misconduct inquiries. If AI tools can reduce the time needed to connect evidence and spot patterns, senior officers will argue that refusing them carries its own public safety cost.
The question is who gets to decide when that cost is acceptable. Police chiefs want operational speed. Mayors and oversight bodies want legality, competition and public confidence. AI suppliers want predictable procurement rules. Citizens want safety without feeling that opaque systems are being quietly embedded into institutions that already hold enormous power over them.
That tension is now the market. Palantir's legal threat signals that major AI vendors may no longer accept political vetoes quietly, especially when large contracts and reputations are at stake. At the same time, Khan's move shows that values-based objections can become a practical tool for stopping or delaying AI surveillance and defence-adjacent deals.
The next thing to watch is whether this dispute produces a fresh procurement process, a court fight or a negotiated compromise that lets the Met buy similar technology under tighter conditions. Whatever happens, the message for govtech is clear. In public AI, the buyer is not only the department signing the contract. It is also the politician, the watchdog, the campaign group and the public whose trust can make the deal possible, or make it impossible.
Also read: Kalshi is making traders disclose jobs to police insider bets • TSMC says AI chip demand will keep prices under pressure • Super Micro turns to Wall Street to keep up with AI demand
