From 28th to 30th April, the STAI CDT hosted a three-day research sandpit focused on advancing AI safety technology in the communications sector. The event brought together STAI CDT PhD students and experts to explore pressing challenges and generate innovative, collaborative solutions.
AI safety technology encompasses the tools and techniques designed to ensure AI systems are safe, secure and aligned with human values. Despite rapid advances in this field, much of the emerging research has yet to be applied or tested in sectors that greatly need it.
To address this gap, second and third-year STAI CDT PhD students worked in self-organised teams to develop research proposals targeting key challenges in the communications landscape. The sandpit began with contributions from a range of experts, who highlighted research gaps and real-world issues facing the sector. Building on these insights, students spent two days collaboratively designing and refining their project ideas.
The challenges taken on by the groups were diverse, with all teams incorporating considerations of equality, diversity and inclusion (EDI) and responsible innovation throughout, while also assessing how their ideas could be developed into viable project proposals.
Several groups concentrated on developing project ideas designed to mitigate the challenges experienced by vulnerable individuals, especially young people, in relation to chatbot use, as well as echo chambers and the ‘filter bubbles’ generated by algorithmic systems. Another group developed a project idea aimed at reducing safety risks associated with the compression process, which occurs when AI applications are adapted for mobile use, often resulting in altered functionality or safeguards.
The sandpit culminated in a series of presentations, where teams pitched their proposals to a panel of judges. The panel included a representative from Ofcom, the UK’s regulator for communications services, who provided valuable feedback and a sector-specific perspective.
The event showcased the creativity and interdisciplinary strengths of STAI CDT PhD researchers, while fostering meaningful engagement with stakeholders and supporting the development of impactful, application-driven AI safety research.