Sadiq Khan has halted a proposed £50 million deal that would give the Metropolitan Police access to Palantir’s data‑analysis platform. The move reflects growing unease about vendor lock‑in, privacy, and the rush to adopt AI tools in public safety, while supporters argue the technology could modernise investigations.
A high‑profile veto on AI‑driven policing
London’s mayor, Sadiq Khan, announced that the Metropolitan Police will not proceed with a £50 million contract to acquire Palantir’s data‑integration and analytics suite. The decision was framed as a safeguard against vendor lock‑in – the risk that the Met would become dependent on a single, US‑based technology provider for critical intelligence work.
The proposed system, often described as an “AI‑enabled” platform, promised to automate the aggregation of disparate data sources – from CCTV feeds to public records – and surface patterns that could aid investigations. Palantir, a company that grew out of the US intelligence community, has long marketed its software to law‑enforcement agencies worldwide, positioning itself as a turnkey solution for “smart policing”.
Signals that the deal was more than a routine procurement
Escalating budgets for AI in public safety – Across the UK, police forces have collectively earmarked over £200 million for AI‑related tools in the last three years. The Met’s £50 million request was one of the largest single‑force allocations, indicating a shift from pilot projects to full‑scale deployment.
Political pressure from civil‑rights groups – Organisations such as Liberty and Big Brother Watch have repeatedly warned that opaque AI systems can amplify surveillance and erode privacy. Their lobbying intensified after a 2023 report highlighted how Palantir’s software was used in US cities to predict crime hotspots, sometimes with biased outcomes.
Internal dissent within the Met – Leaked minutes from a senior officers’ briefing showed a split: some senior analysts praised Palantir’s ability to reduce manual data‑correlation time, while others warned that the platform’s “black‑box” decision‑making could clash with the Met’s accountability frameworks.
International scrutiny – The UK’s National Cyber Security Centre recently issued guidance urging public bodies to assess the geopolitical implications of sourcing critical software from foreign vendors, especially those with close ties to intelligence agencies.
Counter‑perspectives: why some officials still back the technology
Potential operational gains
Proponents argue that Palantir’s platform could cut the time analysts spend stitching together data sets from weeks to hours. In a pilot run with a smaller UK police force, the software reportedly identified a network of vehicle thefts by correlating registration data with social‑media posts, leading to a rapid bust.
Existing contracts and sunk costs
The Met has already invested in data‑integration tools that are compatible with Palantir’s APIs. Critics of the mayor’s veto note that abandoning the project now could waste the £10 million already spent on licences, training, and infrastructure upgrades.
Competitive pressure from other forces
Several UK police forces – including Greater Manchester and West Midlands – have signed contracts with Palantir or similar vendors. If the Met steps back, it may fall behind in the ability to share cross‑regional intelligence, a concern for a city that often serves as a hub for national investigations.
The broader debate on AI procurement in the public sector
The episode sits at the intersection of three recurring tensions:
- Speed versus scrutiny – Governments are eager to adopt AI to keep pace with sophisticated criminal networks, yet the procurement process for high‑risk software often lags behind the rapid evolution of the technology.
- Domestic capability versus foreign dependence – The UK’s own AI industry is still maturing. Relying on US firms like Palantir can deliver immediate capability but raises questions about data sovereignty and strategic autonomy.
- Transparency versus effectiveness – Tools that operate as “black boxes” can deliver results, but they also make it harder for oversight bodies to assess bias, error rates, or compliance with data‑protection law.
What comes next?
Mayor Khan has asked the Met to explore alternative solutions that meet the city’s privacy standards and provide a clear exit strategy. Potential paths include:
- Open‑source analytics stacks – Projects such as Apache Superset and OpenSearch can be self‑hosted, giving the Met full control over data pipelines.
- Domestic AI vendors – Companies like Graphcore and DeepMind (via its safety‑focused research arm) are courting public‑sector contracts, promising tighter alignment with UK regulations.
- Hybrid approaches – Combining bespoke, in‑house analytics for sensitive data with third‑party services for less critical workloads could balance risk and capability.
The decision also signals to other UK authorities that political oversight of AI procurement will remain a live issue. Whether the Met ultimately adopts a different vendor, builds its own platform, or scales back its AI ambitions will likely influence how other municipalities frame their own technology strategies.
The story continues to evolve as the Metropolitan Police releases further details on the cancelled contract and as civil‑society groups prepare formal challenges to any future AI‑driven policing initiatives.
Comments
Please log in or register to join the discussion