As the public sector grapples with the rise of generative AI (gen AI), a panel of experts gathered at the Think Data for Government event in London to discuss the critical skills needed to harness the power of these new technologies while addressing pressing security challenges.

The event brought together a diverse group of leaders, including Eoin Mulgrew, head of innovation at 10 Downing Street, Sherlock di Schiavi, head of security architecture at the Office of Nuclear Regulation, Dave Beech, senior solutions architect at Elastic, and Jessica Figueras, chair of Trustees at the UK Cyber Security Council.
Di Schiavi emphasised the need for transparency and understanding the limitations of artificial intelligence, especially in security-sensitive roles. He also highlighted the importance of quantitative risk assessment to understand the threats posed by AI and whether they can be reduced through externalisation.
Figueras noted the rapidly evolving nature of AI and related fields. “The challenge often…is, what skills are we talking about? It’s a fundamentally very, very fast moving, evolving field.”
Figueras said the required skills in data and cybersecurity have shifted over time, from basic programming knowledge to a broader understanding of data governance, ethics, and risk management.
“Gen AI, again, an entirely new field. I observed that that seems to span from those who are actually developing models to those who describe themselves as prompt engineers,”
If you liked this content…
Upskilling rather than relying on new hires
Elsewhere, Beech emphasised the importance of users understanding the limitations of AI tools to avoid being misled by their outputs, as well as the potential security risks if these systems are infiltrated by bad actors.
In response to the skills gap, Mulgrew, called for a focus on upskilling the existing public sector workforce rather than solely relying on new hires.
“We don’t necessarily need to upskill a load of our workforce to train large language models. The same probably goes for security. But what we do need to focus on, and we can get a lot done, make sure that a large and ever increasing chunk of our workforce has a good solid grounding in the foundations of basic machine learning models how to apply them. Python is still really useful and basic data science and maybe even a little bit of engineering as well.”
Mulgrew shared the success of in-person upskilling programmes, which have delivered 24,000 hours of training to more than1,000 civil servants.
The panel also discussed the importance of leadership and literacy in the public sector. As Figueras explained, “One of the most important things that you need to learn in your leadership journey is how to apply the right lens to the right problem. And it may well be that you know the field you’ve come from, the particular problem in front of you right now that is not a particularly useful lens, and you might have to borrow from other fields.”
The consensus among the panellists was that by investing in upskilling programs, fostering a culture of flexibility and continuous learning, and ensuring leaders have a strong understanding of these emerging technologies, the public sector can better harness the power of AI while mitigating the associated risks.