Cyber-physical risks pose new challenges to the UK security and defence community. Nation state actors are increasingly equipping themselves with tools capable of causing damage to cyber-physical systems. During times of elevated threat, e.g. the Ukraine war, it has become common for such tools to be used in the theatre of conflict, but the risk that they will be used outside the affected region also increases, as does the risk that they will fall into the hands of non-state actors and used more widely than against conventional cyber-physical systems. At the same time, society as a whole is developing a growing reliance on social-media, apps, connected IoT devices and both AI and Mixed Reality. By itself, this also has major security implications leading to high volumes of new crimes that need addressing but, when used in combination with cyber-physical attacks, a qualitatively different form of attack is experienced, known as a hybrid threat because of its similarity to the hybrid cyber/kinetic warfare that has emerged as a feature of recent conflicts. Our team will create a cohort of leaders equipped with specialist skills specifically adapted to tomorrow's cyber-physical risks. They will learn how to (1) manage risks that develop jointly across the cyber and physical domains, (2) reduce risks in one domain by intervening in the other domain, (3) understand how criminal groups protect their interest by operating across domains, and (4) pre-empt the propagation of unintended consequences from one domain to the next. These future leaders will also undertake an intense programme of activities designed to nurture key complementary skills. The eXchange, Lead!, Get it Done, and Next Step schemes will be shaped with our industry partners to address important gaps highlighted in a recent government report: "40% of businesses reported that cybersecurity job applicants are deficient in their complementary skills". Beyond fostering technical expertise, the CDT will help students and partners develop an acute awareness of the wider issues and dilemmas posed by all such work in democratic societies. To deliver an exceptional learning experience, we bring a powerful group of industry partners and academic experts from the engineering, natural and social sciences within a single-site research training centre in London. Academically excellent and pro-active within and between cohorts, the 50+ CDT students will specialise in a range of scientific techniques from at least two of the following four 'cyber-physical risk' themes. These were selected in consultation with students and partners based on their relevance, attractiveness and employment opportunities: 1. Futures: forecast how future socio-technical trends will shape the geopolitical implications of cyber-physical risk, understand how to co-design effective control and mitigation measures for different legal, technical, and policy contexts in order to support societal resilience. 2. Cyber-physical systems: support regulation of IoT devices, detect 'below the radar' attacks against cyber-physical control systems; contribute to the resilience of critical infrastructures. 3. Online Communication: prevent the misuse of social media for disinformation campaigns, detect incitement to hate crime in games, analyse crime related communication on the dark web, etc. 4. Simulation and Interaction: develop and apply simulation technologies (e.g., digital twins, VR, XR) for the study of human behaviour in a range of cyber-physical risk scenarios. EPSRC areas directly relevant to the following themes: Artificial intelligence technologies, Building a secure and resilient world, Digital security by design, Digital signal processing, Human communication in information and communication technologies, Human-computer interaction, ICT networks and distributed systems, Image and vision computing, Natural language processing, Operational research, Pervasive and ubiquitous computing.

Generative Models are AI models that can generate data. Recently researchers have shown that by training these models on large amounts of data (text data from the internet and images) these models learn to understand the regularities of our text and image world so well that they can generate responses to questions and create new images with surprising fidelity. This heralds a new era in which computers can assist humans to carry out tasks more efficiently than ever with significant opportunities for society, science and industry. However, these advances need significant research still -- how to make them train efficiently on different problems, how to understand their reliability and adherence to ethical norms.