UK government should embrace open source for AI – Red Hat

As the UK government ramps up its adoption of artificial intelligence (AI) across the public sector, Jonny Williams, chief digital advisor for the UK public sector at Red Hat, is calling for a shift in strategy toward open source software.

Williams believes the path to trustworthy, efficient and ethical AI in government lies not in proprietary ‘black box’ solutions, but in transparent, open platforms.

“Right now, we’re seeing government departments lean heavily on proprietary AI – black box systems that are monolithic and closed,” said Williams. “There’s little to no visibility into what data these models are trained on, how they reach conclusions, or how bias may be built in. That’s a real concern when you’re making policy decisions that affect real people.”

One of Williams’ most pressing concerns is the risk that AI, as currently implemented, could undermine public trust. “We are talking about the first type of technology introduced into government that is unauditable. There’s no clear audit trail for how decisions are made. That’s deeply troubling in a democratic society.”

The FOI timebomb

Williams also warned of a looming administrative burden: the potential weaponisation of Freedom of Information (FOI) requests. “If a member of the public asks for the logic or code behind a decision – say, how their council tax was calculated – there’s currently no way to provide that when AI is involved. Civil servants will spend huge amounts of time chasing opaque answers. It’s unsustainable and expensive.”

The solution, he said, is open source AI, where models, training data, and codebases are fully visible, shareable and explainable. “With open source, we can actually answer these questions. We can show people how decisions were made, and we can stand behind the technology.”

Williams also said that AI models are often trained on datasets that reflect different values to those of UK institutions and society. “AI trained in American contexts might have a completely different approach to issues like school safety, which could skew policy decisions in the UK,” he explained.

“At a time when trust in public services is critical, using closed AI systems undermines that. People deserve to know how these systems work, especially when they influence decisions in healthcare, education or benefits.”

Why isn’t government already doing this?

The primary barrier to open source implementation is awareness, said Williams.

“There’s nothing technically stopping government from adopting open source AI. The challenge is that senior decision-makers often don’t fully understand the implications of using closed systems – or the benefits of going open.”

He points to the proliferation of proprietary pilots across 467 government departments, many of which may never scale. “We’re suffering from ‘pilotitis’ – a lot of experimentation, but very little that’s production-ready. That’s expensive, inefficient, and ultimately a missed opportunity.”

Williams draws inspiration from past successes like GOV.UK, which consolidated thousands of public-facing websites into one trusted platform. “That worked because it was open, collaborative, and transparent. AI should follow the same model.”

Using AI ethically

Williams pointed to past successes with Ministry of Defence’s D2X platform and the Department for Work and Pensions’ automation initiatives during Covid-19, both built on Red Hat technologies. “These projects prove the value of open source platforms in critical national infrastructure. The next step is embedding open source AI into these frameworks.”

He also references VLM (a high-performance open-source AI project), which offers 24x throughput compared to traditional models. “Open source isn’t just transparent—it’s faster, more efficient, and significantly cheaper to run.”

Williams said he isn’t advocating for the abandonment of AI, but instead adopting a more thoughtful, mission-driven approach. “We’re not saying don’t use AI. We’re saying use it ethically, use it transparently, and use it wisely. Open source enables all of that.

“The way the state works should reflect the values of the nation,” said Williams. “We need AI platforms that are trustworthy, scalable, and aligned with our democratic principles. Open source is the best way to get there.”