AI policy leaders' series: in conversation with Sophie England, Head of UK Public Policy - AI and wearable at Meta

We talk to Sophie about what it’s like to work on some of society’s most interesting issues, what the UK government is getting right and the potential of open-source AI.

Working on society’s most interesting issues

I originally joined Meta to work on some of society’s most interesting and complex issues. That started with social media policy, looking at questions about what should and shouldn’t be said online, including misinformation, hate speech and other forms of harmful content. These are the kinds of issues that people talk about at the dinner table.

AI feels like the next, much bigger, version of that. Social media can be a discrete part of life where you can choose to opt in or opt out. AI is different as it’s becoming embedded across everything we do. It represents a far more fundamental shift and that’s part of what makes it both exciting and difficult to get your head around.

What should AI be used for, and what shouldn’t it? What are the rules of the road? Much of that hasn’t been fully thought through yet, so having the opportunity to work on these issues is a huge privilege. While tech companies often receive criticism, it’s also a privilege to work on these questions from the inside.

A lack of consistency is holding back UK AI policy

The UK government is doing a great job in many respects. The Opportunities Action Plan was a strong start, and in some ways, it’s ahead of what other countries have been doing. Many of the ideas being put forward are practical and achievable. There is real investment and a genuine effort to shift the needle. The Government are also taking a pragmatic approach to AI safety and has avoided introducing a large, overarching AI bill.

However, there is a growing sense of inconsistency. On the one hand, the Prime Minister and Chancellor are pro-technology and emphasise its potential.

But on the other hand, Parliament is not fully aligned, and a significant number of MPs are unsure about AI and are not yet convinced. That is starting to create friction and risks slowing progress.

AI is moving quickly, with new capabilities emerging every few weeks. This creates the risk of a disconnect between the speed of development and the conversations happening in Parliament.

The opportunities and challenges of open-source AI

Open source is an important part of the AI ecosystem. It took some time for the Government to fully engage with it, but it is now gaining traction, particularly in certain pockets such as the AI incubator and among key Ministers.

There is significant potential for open-source AI to boost economic impact. Open-source models can lower barriers to entry and help businesses build and innovate not just in the UK but in the Global South. It also has implications for sovereignty where governments do not need to be reliant on one or two closed model providers.

However, there are also challenges. Some of the most capable open-source models are coming from China, and there are also concerns about security as open-source models become more capable.

There are also practical barriers to adoption. Open-source models typically require more technical expertise, access to compute and the ability to run and maintain systems directly rather than through APIs. That makes them less immediately accessible than closed models.

Procurement is holding back greater adoption

Government procurement processes aren’t as well-suited to open-source models. Open-source models are free to use, but there are costs elsewhere. You need engineers to implement the systems.

However, governments typically don’t have this capacity, and open-source developers can’t provide it because they don’t have go-to-market teams like closed model developers do. As a result, closed-model developers, with full go-to-market teams, are often better positioned to compete for government contracts.

This is why the open-source programme came about, where Meta granted funding to The Alan Turing Institute to find 7 fellows to be seconded into Government to build open-source tools for the Government’s highest-priority challenges. Together, we created a programme that was about upskilling from the inside, rather than an off-the-shelf solution you can buy.

The programme has really caught fire, and it’s going to be interesting to see how this helps unlock the potential of open-source AI across the public sector and the wider economy.