The United Kingdom has decided to keep using Palantir’s NHS Federated Data Platform until its current contract ends in March 2029, rather than terminate it early. Health minister James Frith informed Parliament on 16 November that the two-year extension is “a bridge, not the destination,” giving the government time to seek a new supplier ([1]) ([2]). The government plans to launch a re-procurement in 2027 with requirements for “interoperability, resilience and trusted common standards” to encourage more “diverse and innovative” bidders, including U.K. firms, and avoid dependence on any single foreign vendor ([3]) ([4]).
Palantir’s role as the NHS’s data platform provider has been controversial. The U.S. company’s technology was initially adopted during the pandemic, and it later won a five-year, £330 million contract to unify patient records across the health service ([5]). However, Palantir’s involvement drew “considerable scrutiny from campaigners, health professionals and MPs” because of concerns over data security, ethics and vendor lock-in ([6]). Medical groups argue a publicly funded health system should use a “trustworthy alternative” aligned with NHS values ([7]). In response, officials have promised that strong transparency and data safeguards will be built into the next system. By extending the current deal but committing to an open tender, the U.K. government is aiming to maintain continuity of services while ultimately shifting to a solution with greater public trust and national data sovereignty.
Global information firm Thomson Reuters has highlighted how improving data quality and governance can accelerate AI adoption. The company recently detailed how it built an internal analytics platform, DeepVista, by unifying its finance data into “one consistent version of the truth” on Microsoft’s Fabric cloud platform ([1]). This “semantic” data layer ensures that every metric is defined and governed in one place, so employees across departments see the same numbers when making decisions.
The payoff has been striking. Since launching the new system in June, DeepVista grew from zero to “thousands of active users who have asked thousands of questions” in Thomson Reuters’ Monthly Business Review process ([2]). Meetings that once began with debates over whose figures were correct now start with “everyone open[ing] the same app” where “every figure traces back to its source” ([3]). According to Thomson Reuters’ data and analytics leaders, people “kept coming back because they trusted the answers” their new AI assistant provided, instead of relying on teams of analysts to validate data ([4]). The case illustrates how a foundation of reliable, well-governed data can turn AI from a buzzword into a practical tool for faster, more confident decision-making.
The importance of data governance and ownership was underscored by a landmark legal settlement involving AI startup Anthropic. In July, a U.S. judge approved a $1.5 billion class-action settlement after Anthropic was sued for using hundreds of thousands of pirated books to train its generative AI model, Claude ([1]) ([2]). The lawsuit revealed that Anthropic had downloaded over 7 million copyrighted books from pirate websites without permission ([3]). While the court ruled that training AI on legally obtained text can qualify as fair use, it found that the company’s method of acquiring data was unlawful, leading to what plaintiffs called the “largest known copyright recovery in history” ([4]).
Under the deal, which is now entering its payout phase, authors and publishers will receive roughly $3,000 for each affected title, shared according to who holds the rights ([5]) ([6]). Yet even distributing the funds has proven challenging. Many writers report that publishers have claimed 100% of books where rights have long since reverted to the authors ([7]). The Authors Guild and other advocates attribute the confusion to “bad recordkeeping” and outdated rights data, rather than deliberate malice by publishers ([8]) ([9]). The case stands as a warning that unclear data ownership and poor governance can create not only public trust issues but also tangible financial risks: companies that build AI on data without proper rights or provenance may face regulatory backlash and expensive legal consequences.