DSIT backs open-source AI with compute and mentoring

TL;DR:

• The Department for Science, Innovation and Technology is giving open-source AI builders over £500,000 of compute — 160,000 GPU-hours from the AI Research Resource.
• The funding goes to the strongest projects from the NVIDIA-backed Hack for Impact hackathon, alongside a new mentoring scheme and a developer advisory board.
• A RIBA-partnered data-centre design challenge and fresh robotics safety guidance round out the package.

The UK government is putting weight behind open-source artificial intelligence, with AI Minister Kanishka Narayan announcing compute, mentoring and a developer board aimed at turning prototypes into public-service tools. The pitch carries a strategic subtext: open models as a route to cut the public sector’s dependence on US vendors.

From hackathon to public services

The headline measure is more than £500,000 of processing power — 160,000 GPU-hours drawn from the AI Research Resource — awarded to the best entries from the recent Hack for Impact event, run with chipmaker NVIDIA. Winners range from Codeborough, which pulls scattered civic data together to help Londoners find libraries, toilets and polling stations, to projects spotting gaps in NHS waiting-list care and keeping 999 services running over satellite during mobile outages.

Two support structures sit alongside the cash. An Open-Source AI Builder Mentoring Scheme pairs winners with experts from the government’s Incubator for Artificial Intelligence, while an Open-Source AI Dev Board gives ten UK developers under 30 a direct line into government through ministerial roundtables across 2026. The government is also launching the RIBA x DSIT Data Centre Design Challenge, asking architects and engineers to rethink data centres around sustainability and community value — and has struck a regulatory partnership with the Health and Safety Executive on guidance for robots working safely alongside people.

The open-source push complements a run of UK sovereignty moves, from the £1.1bn plan to back British AI chips to the new Cambridge sovereign-AI lab with AMD and Dell.

Looking forward

The sums are modest next to private capital, but the direction is deliberate: nurture homegrown, open AI talent and reduce reliance on a handful of foreign platforms. Whether £500,000 of compute and ten advisory seats translate into deployable public-service tools — rather than another hackathon afterglow — will be the measure of intent.