Foreign-built AI models are now drafting British legislation, shaping defence analysis, and influencing education policy — and most business leaders have no idea it’s happening. A New Statesman investigation reveals how US and Chinese technology companies have quietly embedded their systems into Westminster’s decision-making machinery, bypassing democratic oversight entirely. For UK organisations that depend on government contracts, regulatory frameworks, or public sector partnerships, this isn’t an abstract policy concern. It’s an immediate strategic risk.
The governance gap nobody’s talking about
The scale of AI adoption across UK government has outpaced any attempt to regulate it. According to the New Statesman’s reporting, the 2025 Spending Review mentioned AI 38 times and committed £2 billion to AI development. But the infrastructure for overseeing how these systems actually work — who built them, what biases they carry, whose interests they serve — barely exists.
Strategic Reality: The UK government has spent £476 million on AI consultancy contracts since 2022 while simultaneously failing to build internal technical expertise. Organisations dealing with government face an institution that’s increasingly AI-dependent but structurally unable to evaluate what these systems are doing.
The problem isn’t that government is using AI. The problem is that multiple underfunded agencies — the Department for Science, Innovation and Technology, the Government Digital Service, the Office for AI — are competing over oversight responsibilities without any of them possessing genuine regulatory authority or deep technical expertise. The result is a regulatory vacuum filled by the very companies being regulated.
| Factor | Current state | Risk level |
|---|---|---|
| Internal AI expertise | Fragmented across competing agencies | High |
| Regulatory authority | No single body with enforcement power | Critical |
| Democratic scrutiny | AI in legislation without disclosure | Critical |
| Vendor independence | Heavy reliance on US/Chinese models | High |
| Financial oversight | £476m in consultancy with limited accountability | Medium |
What’s actually happening inside government
Will Dunn’s investigation paints a picture that should concern any organisation operating in the UK. AI-generated text has entered parliamentary legislation. Systems built by foreign companies influence policy analysis across defence and education. And there’s a revolving door between the technology industry and the bodies meant to oversee it.
Critical Context: Sixty House of Lords members have declared interests in AI companies. This isn’t a peripheral conflict of interest — it’s a structural one that shapes the laws governing AI use across every sector.
The economic narrative driving this adoption deserves scrutiny. Officials cite projections of £37–45 billion in annual savings from AI. But as the investigation notes, these figures depend on radical governmental restructuring, not simple efficiency gains. The gap between projected savings and actual implementation reality matters because it shapes the procurement decisions, regulatory frameworks, and public sector partnerships that affect every UK business.
Research cited in the article shows that chatbot systems employ persuasion techniques and sometimes generate false information. System prompts — the hidden instructions that shape AI behaviour — can embed political biases reflecting their creators’ ideologies. When these systems inform government policy, those biases flow directly into decisions affecting businesses, communities, and individuals across the country.
Hidden Cost: The £2 billion allocated for AI development creates a powerful incentive structure. Government departments competing for this funding are motivated to adopt AI quickly rather than carefully — prioritising speed over the governance frameworks that protect public and business interests.
The strategic picture for UK organisations
This situation creates three distinct categories of risk for UK organisations, and most leadership teams aren’t tracking any of them.
Regulatory unpredictability. When the regulator lacks technical expertise and the regulated industry has captured the oversight process, regulatory outcomes become unpredictable. Organisations planning compliance strategies around AI governance face a moving target shaped more by industry lobbying than coherent policy design.
Procurement dependency. Government’s increasing reliance on a small number of foreign AI providers creates concentration risk that extends to every organisation in the public sector supply chain. If government systems shift platforms, change procurement frameworks, or face geopolitical restrictions on technology access, the ripple effects hit contractors and partners immediately.
Trust and accountability gaps. When AI-assisted government decisions affect planning permissions, regulatory approvals, or public sector contracts, organisations need to understand what influenced those decisions. Right now, there’s no transparency requirement and no clear accountability framework.
Strategic Insight: As one government technology adviser told the New Statesman, “This is a war” between engineers, the wealthy individuals controlling AI models, and politicians seeking authority. UK organisations are caught in the crossfire of a power struggle they didn’t choose and can’t directly influence.
| Stakeholder | Primary concern | Action required |
|---|---|---|
| Board and C-suite | Regulatory uncertainty affecting strategic planning | Scenario planning for multiple regulatory outcomes |
| Procurement teams | Vendor concentration risk in supply chains | Map AI dependencies in government contracts |
| Compliance officers | Shifting accountability frameworks | Build adaptive compliance processes |
| Public affairs | Opaque government AI decision-making | Establish transparency monitoring capabilities |
| Technology leaders | Foreign model dependencies in UK infrastructure | Evaluate sovereign AI alternatives |
What UK organisations should do now
The temptation is to wait for clearer regulation. That’s the wrong approach. The governance vacuum described in this investigation means organisations need to build their own frameworks now rather than relying on government to provide one.
Implementation Note: Organisations don’t need to solve all of this at once. The priority actions below are sequenced by maturity level — start where you are, not where you think you should be.
For organisations starting their AI governance journey:
- Audit your current exposure to government AI systems — map every touchpoint where AI-influenced government decisions affect your operations
- Document which foreign-built AI models your organisation uses directly, and which are embedded in your supply chain
- Establish a board-level AI governance brief that includes public sector AI risk alongside internal AI deployment
For organisations with existing governance frameworks:
- Extend your AI risk assessment to include regulatory capture risk — the likelihood that governance standards will be shaped by vendor interests rather than public benefit
- Build scenario plans for three regulatory futures: light-touch (industry self-regulation wins), prescriptive (EU-style regulation), and fragmented (competing agency mandates)
- Invest in relationships with the multiple government bodies overseeing AI — DSIT, GDS, the Office for AI — rather than assuming a single point of contact
For mature organisations leading on AI governance:
- Consider contributing to open-source or UK-sovereign AI alternatives that reduce dependency on foreign-controlled models
- Develop transparency standards for AI-influenced government decisions that affect your sector, then advocate for their adoption
- Share your governance frameworks with industry peers — collective standards carry more weight than individual company policies
Four risks hiding behind the headlines
The skills drain nobody measures. Government’s £476 million spend on AI consultancy isn’t just expensive — it actively prevents internal capability building. Every consultant engagement is a missed opportunity for civil servants to develop expertise. This creates a permanent dependency loop that keeps government reliant on external advice and unable to challenge vendor claims independently.
Resource Reality: The skills gap compounds over time. Each year without investment in internal government AI expertise widens the gap between what officials can evaluate and what vendors can sell them. Organisations should factor this growing asymmetry into their risk models.
Geopolitical fragility in model supply chains. UK government’s reliance on US and Chinese AI models creates a vulnerability that has no precedent in British public administration. Trade disputes, sanctions, or technology export controls could restrict access to the very systems now embedded in defence analysis and policy-making. Organisations in the public sector supply chain inherit this fragility.
The lobbying feedback loop. With 60 House of Lords members declaring AI company interests and a revolving door between industry and oversight bodies, the risk isn’t just biased regulation — it’s that regulation becomes a mechanism for market incumbents to raise barriers against competitors. Smaller UK AI companies and their business customers face a playing field tilted toward whoever has the strongest Westminster relationships.
Democratic legitimacy risk. When AI-generated text enters legislation without disclosure, the democratic basis for those laws is compromised. For regulated industries, this creates a novel compliance question: are you obligated to comply with laws that were partly written by systems operating outside democratic accountability?
Warning ⚠️: Organisations that treat this as purely a technology issue are misreading the situation. This is a governance and power question with technology at its centre. The response needs to come from boards and leadership teams, not just IT departments.
Where this leaves UK organisations
The core finding from the New Statesman’s investigation is straightforward: foreign-controlled AI systems have been embedded into British government faster than any oversight mechanism can track. The economic promises justifying this adoption don’t hold up under scrutiny, and the governance structures meant to manage it are fragmented, underfunded, and captured by the industry they’re supposed to oversee.
For UK organisations, three things matter most:
- Map your exposure now. Every organisation interacting with government needs to understand where AI-influenced decisions affect their operations, contracts, and regulatory obligations
- Build governance before you’re told to. The regulatory vacuum won’t last forever, but organisations that build robust AI governance frameworks now will be better positioned when regulation does arrive — whether it’s light-touch, prescriptive, or somewhere in between
- Engage with the process. The investigation describes a system where technology companies dominate the conversation. UK organisations have both the right and the business interest to demand transparency, accountability, and genuine democratic oversight of AI in government
Take Action: Start with a simple exercise — list every government decision that affects your organisation, then assess which of those decisions might now involve AI systems. That map is your starting point for building a response to this emerging risk.
This analysis is based on “The Silent Coup: How AI Captured Westminster” by Will Dunn, published in the New Statesman on 9 April 2026. Resultsense provides UK-focused AI analysis to help organisations make sense of artificial intelligence governance and adoption.