Summary

• DSIT has announced compute, mentoring, and a developer board for open source AI builders.
• The package includes more than £500,000 worth of compute and 160,000 GPU-hours from the UK AI Research Resource.
• The policy links AI adoption to public-service tools, infrastructure design, robotics guidance, and domestic developer capacity.

The Department for Science, Innovation and Technology has announced support for open source AI developers, including compute access, mentoring, and a new developer board intended to give younger UK builders a route into government policy discussions.

AI minister Kanishka Narayan set out the measures during London Tech Week, following a DSIT and NVIDIA Hack for Impact event where developers used open data from the City of London to build tools for public-service and city-infrastructure problems.

The government’s official update includes an Open-Source AI Builder Fund offering more than £500,000 worth of compute. That will include 160,000 GPU-hours from the UK AI Research Resource, aimed at helping projects move from prototypes to public AI tools.

DSIT is also creating an Open-Source AI Builder Mentoring Scheme, pairing hackathon winners with experts from i.AI, the government’s in-house artificial intelligence team. A new Open-Source AI Dev Board will give ten UK-based developers under the age of 30 a channel into government, chaired by Narayan and convening through a series of roundtables in 2026.

The examples given by government are practical: helping Londoners find libraries, toilets, and polling stations; supporting NHS patients on waiting lists by identifying gaps in care and drafting follow-up communications; identifying unclaimed business grants; warning of roadworks and transport disruption; and using satellite connectivity to keep emergency services running when mobile networks fail.

Those examples are more revealing than the funding amount. Half a million pounds of compute will not rebalance the global AI market, although it can give small teams enough capacity to move credible civic tools beyond a demo. The policy signal is that open source is being treated as an adoption route for public-service and civic technology, not only as a developer culture or a counterweight to closed foundation-model providers.

Open source AI gives government a different set of trade-offs. It can support transparency, adaptation, auditability, and local experimentation, but it still requires compute, maintenance, security review, procurement routes, and accountability when tools are used in public services. Releasing or reusing code does not remove the need for service ownership.

The package also includes a RIBA x DSIT Data Centre Design Challenge, intended to explore how data centres can be designed with better public engagement, sustainability, and civic value. That sits awkwardly but usefully alongside open source AI. The software may be open, but the infrastructure that supports AI is physical, capital-intensive, and often locally contested.

Guidance on using robots safely alongside people at work adds a third strand. Taken together, the measures connect AI development with infrastructure, public-service deployment, and workplace automation rather than treating AI as a standalone software boom.

The harder test is whether the programme creates durable routes from promising prototypes into maintained services. Public-sector AI pilots often stall when procurement, legal accountability, integration, security, and operational funding arrive after the prototype rather than before it. The open source route may widen participation, but delivery still depends on turning code into tools that public bodies can safely adopt, support, and improve.