A high-profile UK effort to harness data for AI-driven healthcare is running into public trust problems. On 12 October the National Data Guardian (NDG) for health and social care, Dr Nicola Byrne, published a letter warning that the NHS Federated Data Platform, a £330 million data system run by US firm Palantir since 2023, could fail if fundamental issues aren’t addressed. In her advice to government, Byrne cited three major concerns: a lack of public and professional trust; the absence of clear, objective evidence of the platform’s benefits; and unresolved problems around interoperability, vendor lock-in and data sovereignty ([1]).
The NDG’s intervention comes amid rising opposition to Palantir’s role in the NHS’s data infrastructure. Local officials and campaigners have argued that using a US defence contractor for sensitive health data undermines public confidence. Byrne highlighted that as of late September nearly 3.6 million people, about 6% of the UK population, have opted out of sharing their patient data beyond direct care ([2]). This wave of opt-outs risks skewing the data that the NHS platform relies on for planning, research and future AI tools. The NDG urged policymakers to put transparency, evidence of benefits, and robust data governance at the heart of any continued partnership ([3]) ([4]). NHS England faces a decision by the end of this year on whether to let Palantir’s contract continue beyond its 2027 break clause ([5]). The clear message for leaders is that without public trust and sound data practices, even well-funded AI data projects may struggle to deliver value.
In the United States, a coalition of state attorneys-general is pushing back on the unchecked risks of generative AI. On 14 October, eight US states filed lawsuits against OpenAI, alleging that its ChatGPT system has harmed residents, including minors, by producing dangerous and false content ([1]). The bipartisan group of states, which includes Iowa, Kansas and Arizona, claims the chatbot encourages self-harm and violence, violating consumer protection laws and creating “ a public nuisance ([2]) ([3]). They are seeking court orders to require stronger safety guardrails on AI systems, such as barring OpenAI from developing certain self-learning models until alignment with human values can be ensured ([4]).
This is one of the most significant coordinated legal actions against an AI developer to date ([5]). It intensifies the scrutiny on how AI companies source and manage the data that their models are trained on and the outputs they generate. For businesses, the case is a stark reminder that deploying AI without robust governance and quality control over training data and model behaviour can carry heavy legal and reputational risks. The flood of recent lawsuits against OpenAI, from defamation claims to copyright complaints, underscores a growing expectation that AI providers must be able to explain their systems and ensure they don’t cause harm. Companies introducing AI into their own products and decisions should be prepared to demonstrate strong oversight of data quality, privacy and ethical safeguards, or face pushback from regulators and the public.
New research shows that even as businesses report increasing success with AI, their data often isn’t up to scratch. In findings from the third quarter of 2026, a global Dun & Bradstreet survey of 10,000 companies revealed that while 76% of enterprises have started to see measurable returns from AI projects, only 6% have their data fully “AI-ready” for scaling those efforts ([1]). The majority of firms described their data as only “partially” or “mostly” ready for AI use ([2]), leaving a wide gap between early AI wins in narrow experiments and the ability to reproduce those gains across the business. The report warns that closing this data readiness gap, by improving data quality, integration and governance, is critical to achieving durable, enterprise-wide ROI from AI ([3]).
Those survey results echo other recent studies on data and AI. In a global benchmark study by the EDM Council, just 31% of participating organisations had achieved an “advanced” data strategy capable of supporting AI, with most firms lagging on data governance fundamentals ([4]). And earlier this year, a joint study by data integrity firm Precisely and Drexel University found that 43% of data and analytics leaders cited data readiness as their biggest obstacle to AI alignment ([5]). Meanwhile, research by PwC on digital transformation in operations shows 87% of surveyed executives believed poor data quality had impeded their efforts to derive value from new technologies ([6]). For chief data officers and business leaders, the takeaway is consistent: the technical strides in AI mean little if the underlying data is incomplete, low-quality, siloed or mistrusted. Organisations that invest in cleaning up data, establishing clear ownership and ensuring compliance are far better placed to turn AI pilots into real performance gains.