Introduction: The AI tipping point

The public is increasingly concerned about artificial intelligence (AI). In the UK, AI is perceived as one of the top three risks to humanity, trailing only climate change and war. The capabilities of AI models are advancing rapidly, with models now able to autonomously complete complex software tasks that would normally take a skilled engineer 12 hours. The report warns that capability progress is outpacing safety, labour market impacts are beginning to emerge, and deployment remains uneven.

The UK and EU’s approach to AI is missing direction

Policymakers in the UK and EU are currently caught between accelerating AI deployment and ensuring AI safety. While recent policies have pivoted dramatically towards acceleration—focussing on attracting investment, expanding computing power, and growing domestic champions—they have largely bypassed “AI directionism.” The report argues that growing the AI sector should be a means to an end, not the end itself. Current efforts to build public-value AI remain modest compared to the ambition behind acceleration.

How to implement AI directionism in practice

The report proposes a how-to guide for governments to direct AI towards public value across four key policy areas. First, governments should steer AI towards the highest-impact uses by making bold bets on public-value creation, such as embedding engineers within frontline teams in schools, hospitals, and local government. They should also use outcomes-based procurement to drive transformative outcomes and create markets through regulation. Second, priority sectors must be equipped to adopt AI well. This involves building institutional infrastructure and setting clear standards for responsible AI to enable high-risk, high-reward adoption. Third, power in the AI economy must be rebalanced. Governments should enforce competition policy to curb the entrenched power of incumbent technology giants and use industrial policy to build plural, public-interest alternatives, including open-source AI. Fourth, the gains from advanced AI must be distributed broadly. Recommendations include using fiscal and industrial policy to maintain employment over the medium term by incentivising augmentation over full automation, and giving the public a direct stake in AI’s upside through sovereign AI funds that capture and redistribute windfalls.

Conclusion

European governments require a convincing positive vision for AI to counter growing anti-AI sentiment. Growing the AI sector and hoping for spillover benefits is an incomplete strategy. Governments must become more interventionist by steering AI towards public value missions, preparing sectors for radical transformation, confronting concentrated power, and ensuring the economic benefits are broadly distributed.