The US Government has just published its AI action plan. Given the UK’s own AI Opportunities Action Plan was published earlier this year, and that both plans touch directly or indirectly on education and research, it’s worth comparing the two. What’s changing in the US? How does it differ from the UK approach? And what could this mean for us in UK colleges and universities? These are some of my initial thoughts.
Two AI Plans, Two Worldviews
At first glance, the UK and US plans cover similar ground: compute infrastructure, AI skills development, adoption in the public sector, and a focus on AI safety and evaluation.
But the underlying narratives are very different. Both have three pillars, and those give us a taste of the difference:
UK:
- Lay the foundations to enable AI
- Change Lives by Embracing AI
- Secure the Future with Homegrown AI
US:
- Accelerate AI Innovation
- Build American AI Infrastructure
- Lead in International AI Diplomacy and Security
The UK plan is mission-driven and framed around public benefit. It talks about creating growth zones, stimulating the AI economy, and supporting assurance ecosystems. It includes goals like better public services.
It also talks about the need for the UK to become an ‘AI Maker’ and not and ‘AI Taker’, and need for AI to control its own AI destiny. I’ve talked about AI sovereignty before, and I think the US action plan emphasises the need for the UK to double-down on this.
The US plan talks about competition, dominance, and control. It’s explicit about its aim to “win the AI race,” especially in the face of perceived threats from China. It includes provisions for military adoption, export controls, and changes to the federal risk framework that govern how AI is assessed and regulated. Climate, DEI (diversity, equity and inclusion), and misinformation criteria are out; national security and industrial resilience are in.
The US plan is also clear that, for them, this is about using AI to embed “American Values” worldwide. Interesting they see open source and open weights models as the route to this, noting:
“We need to ensure America has leading open models founded on American values. Open source and open-weight models could become global standards in some areas of business and in academic research worldwide. For that reason, they also have geostrategic value.”
There is of course a contradiction here: they also say:
“AI systems must be free from ideological bias and be designed to pursue objective truth rather than social engineering agendas when users seek factual information or analysis. AI systems are becoming essential tools, profoundly shaping how Americans consume information, but these tools must also be trustworthy.”
Of course, you can never have a purely neutral large language model – it’s always a function of the training date, the fine tuning and so on. But surely it’s a contradiction to say that a model must both be founded on American values and free from ideological bias.
What the Two Plans Have in Common
Despite the tone and politics, there are some genuine areas of overlap between the US and UK plans, and I think it’s worth looking briefly at these.
Compute and Infrastructure Investment
Both plans recognise that scaling AI requires massive compute power and both see this as an area of national interest. The US plan commits to grid upgrades, national data centres, and classified computing environments for defence. The UK’s approach includes AI Research Resource (AIRR) expansion, AI Growth Zones, and a long-term sovereign compute strategy.
A focus on data
Both the UK and US action plans recognise that access to high-quality, well-governed data is essential for training and deploying effective AI systems. The UK plan places strong emphasis on public sector datasets, proposing to mobilise high-value data assets through initiatives like the National Data Library and to reform text and data mining rules to enable AI innovation. Similarly, the US plan calls for the release of more federal datasets in machine-readable formats, the development of shared data platforms, and partnerships with industry and academia to unlock “safe and secure” access to training data.
Workforce and Skills Development
There’s shared urgency around the need to develop AI skills. The US is focusing on workforce retraining, apprenticeships, and integrating AI into blue-collar roles as well as research. The UK wants to train “tens of thousands” of additional professionals and attract global talent, particularly into applied AI fields.
AI in Government
Both plans emphasise public sector transformation. The US focuses and efficiency, and, for example, proposes mandated universal access to LLMs for civil servants. The UK’s approach is similar but less proscriptive:
“AI should become core to how we think about delivering services, transforming citizens’ experiences, and improving productivity.”
Frontier AI and Safety
Both acknowledge the risks of advanced models and propose national institutions to evaluate and mitigate them. In the UK, this means the AI Safety Institute. In the US, risk evaluation is being embedded in existing agencies.
So far, so familiar. But it’s where the plans diverge that things get interesting and perhaps more consequential for the UK’s education sector.
Where the Plans Diverge and Why That Matters
Values and Framing
The UK plan talks about “missions”, public good, and delivering benefits across the economy and society. It’s framed around inclusive growth, responsible innovation, and trust.
The US plan is about competitiveness, national security, and winning. It includes directives to rework existing frameworks, like NIST’s AI risk guidance, to remove references to equity, climate, and misinformation. In their place: adversarial robustness, model alignment, and export controls.
This matters because these aren’t just policy choices. They shape how future AI systems will be built, evaluated, and used. If one country prioritises fairness and the other prioritises control, we’re likely to see very different AI ecosystems emerge.
Regulatory Philosophy
The UK plan tries to tread a middle path. It supports pro-innovation regulation, encourages testbeds and assurance tools, and maintains a flexible approach to compliance.
The US plan is moving in the opposite direction. It’s stripping back regulation that’s seen as ideological or blocking innovation, and aims to limit funding to states where their regulatory framework is seen as excessive.
They do, however, both advocate the use of regulatory sandbox – a way of testing innovative products with real users whilst some of the regulations are relaxed.
Military and National Security
This is where the US plan goes much further. There are entire sections devoted to AI for defence, and indeed the open statement says:
The United States is in a race to achieve global dominance in artificial intelligence (AI). Whoever has the largest AI ecosystem will set global AI standards and reap broad economic and military benefits
It proposes a secure compute environment for national security work and plans to embed AI across the intelligence community. It also focuses heavily on AI interoperability, but not for the public good. It’s so AI can be used in national security.
The UK plan doesn’t directly address military use.
Institutional Development and National Champions
The UK wants to create “UK Sovereign AI” – a new public-private institution to attract investment and build home-grown capability. It also wants to use procurement as a lever to shape the market.
The US is doing something similar, but with more firepower. Its strategy involves alliances with trusted partners, export controls, and incentives to build and keep key AI companies onshore. It’s explicitly about shaping the global market – and keeping adversaries out of it.
And a left turn – Biosecurity
The US Plan ends with a call to invest in Biosecurity. The fear is that AI could “create new pathways for malicious actors to synthesize harmful pathogens and other biomolecules”. This feels like a really curious point to end the plan on! I’m guessing it was put in at the end after lobbying?
What does this mean for UK Universities and Colleges?
It’s hard to overstate this: my reading is that the US plan sees AI models as a means of embedding its values globally, including in academia. The route for this, perhaps slightly surprisingly, is through open source and open weights models developed in the US.
I’ve always argued that open-source/weights AI and open-source software aren’t equivalent when it comes to freedom and control. If you’re interested in exploring that further, I’ve written a blog post on the topic. In short, with open-source LLMs, control over the product still largely rests with the technology provider, as they control the training stage. You also can’t inspect all the code or fully understand how it works in the same way you can with traditional open-source software, so your engagement is really as a consumer, not a potential contributor.
This perhaps reinforces that view – by providing the open-source software, the US government ensures that it retains control over the core of the AI model.
This should be cause for concern. It’s not a new issue, but this is the first time I’ve seen it laid out explicitly as government policy.
In addition, the directive to remove references to diversity and climate change from the NIST AI Risk Management Framework is alarming, especially when the aim is to export US technology to allies. Do we really want AI technologies that aren’t developed by diverse teams? I certainly don’t.
How does the UK’s approach compare? There are some very positive commitments, such as the goal to increase the diversity of the talent pool. But there’s no mention of environmental impact, and a single mentioned of ‘sustainability’. And let’s not forget that the UK has never had an equivalent to the NIST AI Risk Management Framework, a tool that, until this latest US plan, explicitly supported diverse voices in the development and assurance of AI. So there is work to do in the UK too on this.
Will this make international collaboration more complex? Potentially.
But really, my conclusion has to be that I fully support the UK’s plan for our own, homegrown AI solutions, as well as very carefully considered international partnerships.
Find out more by visiting our Artificial Intelligence page to view publications and resources, join us for events and discover what AI has to offer through our range of interactive online demos.
For regular updates from the team sign up to our mailing list.
Get in touch with the team directly at AI@jisc.ac.uk