UK AI ACTION PLAN - timeline
Lots of opportunity in 2025 for UK and AI - don’t get caught napping!
This week the UK’s Department for Science, Innovation & Technology released two key artificial intelligence (AI) documents:
• AI Opportunities Action Plan by Matt Clifford (Plan)
• AI Opportunities Action Plan: government response (Response)
Matt Clifford is a man who ‘gets’ technology and AI he had some bold things to say. This Plan is not crafted by career civil servants and it is well worth reading in full. The Response is encouraging for the sector too. Hat’s off to them.
Clifford makes 50 detailed recommendations in total. Government has “Agreed” to 48 of these, the other two being “Partially agreed” (so that’s good!)
Some key things to watch out for arising out of the Plan and Response as we begin 2025 in earnest:
Availability of data assets for AI training
Very soon we should expect the Department for Media, Culture & Sport (DCMS) and DSIT to start engaging with partner organisations and industry to establish copyright cleared British media assets.
Note that there is currently a separate consultation on Copyright and Artificial Intelligence, specifically addressing “text and data mining” which closes at midnight on 25 February 2025 – likely to be a large number of responses to work through – a possible reason for being one of the two “partially agreed” ones?
Regulation of AI
It’s clear that there will be more AI regulations, but also AI permissions (see above). In the Spending Review in Spring 2025 we will start to see how much money is available for upskilling existing regulators.
We are lucky in the UK to have regulators who are not afraid of technology and are realistic about balancing innovation and regulation. In my view the UK sits emotionally half-way between the EU and the US when it comes to regulation of technology (and what technology is used for).
We will expect more engagement on AI from all core regulators, including ICO, Ofcom, FCA, etc. These will all continue to play a part alongside the enhanced AI Safety Institute (AISI). We will not expect to see our own version of the EU’s AI Act. The UK still does not have an appetite for this sort of regulation, which Clifford clearly thinks of as heavy handed.
Section 1.4 of the plan makes this very clear: “Government must protect UK citizens from the most significant risks presented by AI and foster public trust in the technology … we must do this without blocking the path towards AI’s transformative potential”. In fact, Clifford may view regulators as ‘helpers’ of AI growth (sandboxes, less regulated pilots, etc). Safety regulation unlikely to be eased up.
Compute infrastructure planning and investing:
Expect a long term compute strategy in Spring 2025 with a 10-year roadmap. This will allow builders of AI data centres, and builders of frontier models, to plan.
And if they can plan, investors can plan to invest. Private investment will sit alongside the government’s plan to start the delivery of a new supercomputing facility for AIRR.
AI Growth Zones
The first one is at Culham, the headquarters of the UK Atomic Energy Authority. In Spring 2025 it will be announced how more will be chosen. We will start to see gravity wells of organisations, investment and people around these. (Let the lobbying begin!)
Government contracts
Clear that the government must put it’s money where it’s policy is. Quite how that might pan out isn’t being let out of the bag until Autumn 2025 it seems. Optimistically, it should become easier to sell AI tools/enabled services to the government.
Compliance
There will be no getting away from it, compliance and AI are hard and cover many overlapping areas (data, content, sectoral regulation, employment, safety, etc).
But it looks like over Summer 2025 there will be an AI Knowledge Hub. This is welcome (and should sit neatly alongside the government’s current efforts on assurance tools that don’t over egg the pudding – see for example the government’s AI Management Essentials Tool consultation which closes at 11:55pm on 29 January 2025).
International cooperations
Expect this to be a slower burn as things settle politically in the US and EU.
Incentives to invest in the UK
Unclear right now – likely because Rachel Reeves has no wiggle room on tax incentives. Possibly some opposition to AI being treated differently than any other R&D? Too early to tell.
Education
Clifford has identified a massive need for the UK’s people and organisations to be upskilled. Whilst 2025 may see some announcements here, it’s likely that 2026/2027 will be the year education and AI are really focussed on by the government. It will take time to assess in detail and then plan which bits of the UK educational infrastructure will be support that need.
Visas
Clifford has also identified a massive need for talent to come to the UK. The Department for Business and Trade (DBT) does currently have Global Entrepreneur and Global Talent programmes, but DBT will likely need to address these over 2025 and early 2026. There is definitely some political nervousness around immigration as recommendation 21 related to graduates was one of the two “partially agreed” ones.
Public/private
From Summer 2025 we expect to see the government work hand in hand with the private sector, in a reinforcing way. DSIT has identified and published five ‘quick wins’ to help this (although 5. is a little nebulous!):
Scale and open source 1-2 public sector-led AI solutions that are currently in pilot phase. (For more about what open might/should/could mean in AI see here )
Scale a citizen facing AI tool that enables citizens to engage with government in a more personalised and efficient way
Run Hackathons, aligned to the five key missions. This will be a key way to engage startups in mission delivery.
Pilot the AI Knowledge Hub.
Appoint an AI lead for each mission to help identify where AI could be a solution.
More ‘bodies’ doing stuff:
From Summer 2025 we may well see an expansion of bodies who are tasked with unblocking blockers and really drive AI for the UK. But there will certainly be a new unit to partner with AI companies for AI Growth Zones, National Data Library, etc. Hidden the weeds of the Response, there’s also a recognition that there need to be partnerships between AI companies and the UK’s national security community.
Exciting times.
Lots to do.
And of course, there is already much to do already from a governance point for view for any builder or user of AI in business or any other organisation.
AI governance work starts now.
THANKS FOR READING!