Artificial intelligence presents one of the defining tensions in technology policy today and is characterised by the pull between moving fast to remain competitive with allies and adversaries alike, and moving carefully enough to get the fundamentals right.

Generative AI is different to most regulatory challenges because it is a technology which could be dangerous in itself, not just in the wrong hands.

Regulation, in this context, is not a single lever but a series of judgement calls - on safety, on copyright and on how much power sits with the state versus the developers of frontier models.

I have met organisations across the spectrum of opinion on this, from those who urge caution about the most extreme risks AI may pose, to those who are optimistic about its potential to transform our economy and public services. We should pay careful attention to both perspectives.

This week, as part of Evidence Week in Parliament, I spoke with researchers from the Tony Blair Institute, who set out how AI could help government test and deliver evidence-based policy, moving public services from reactive to preventative, and from standardised to personalised.

I also met experts from the Oxford Martin AI Governance Initiative, whose work focuses on how the UK can secure sovereign access to frontier AI, a question of national resilience as much as economic opportunity.

These topics are a longstanding interest of mine. As Secretary of State for Digital, Culture, Media and Sport, I spoke about the need for both a skilled workforce and clear ethical foundations if Britain was to lead in AI. Outside Government, I went on to chair the All-Party Parliamentary Group on Digital Regulation and Responsibility from 2019 to 2024, and the questions we grappled with then remain just as live today.

Part of getting this right means ensuring parliament itself keeps up. There are now structured training pathways available to staff here, ranging from basic grounding in what generative AI is and how it works, through to more advanced modules on prompt engineering and critical thinking when using these tools.

Download our Mobile App!

Equipping our own institution with this literacy is a small but necessary step in making sure our democracy is not left behind by the technology it is being asked to regulate.

There is also a useful parallel here with social media regulation. As Secretary of State, I introduced the Online Harms White Paper in April 2019, setting out the case for a statutory duty of care on tech platforms. It took until October 2023 for the Online Safety Act to receive Royal Assent, and full implementation has continued to stretch out since, with key duties still being phased in as we head through 2026.

That is the better part of seven years from first principle to practical effect. AI regulation cannot afford so slow a path since the technology moves faster, and the risks of getting the sequencing wrong, whether through inaction or overreach, are more significant.

Still the example of the Online Safety Act shows the value of getting the detail right, rather than rushing to legislate for its own sake.

The task ahead for government is not whether to recognise that AI will change our public services and economy (that much is inevitable) but how it chooses to shape that change.

Done well, AI can make services more responsive and government more effective. Done carelessly, it risks entrenching huge new problems as fast as it solves old ones.

Getting this balance right is the work now facing us as legislators.