Topics like ransomware, hacktivism, offensive cyber operations, and the AI supply chain live in an educational grey zone. They are often considered too technical for the core political science syllabus, yet too socio-political for software engineering degrees. Understanding how technical, social, and political dimensions interact, however, is no longer optional.
That’s exactly why we launched the 12-week ‘Foundations of Cybersecurity and AI’ online live course back in April: to create a virtual classroom for a multidisciplinary cohort of students, with or without a technical background, to engage holistically with these issues.
The gap filled by our course
Artificial intelligence challenges us to rethink how systems, including critical infrastructure, are attacked and defended. Meanwhile, cybersecurity concerns influence how AI is developed and deployed. Today’s political decisions around related regulation establish benchmarks for our increasingly digital environment. But these benchmarks are only as good as the individuals who understand the underlying trade-offs.
Equipping students with the knowledge and tools to form a critical understanding is the rationale behind our course. Over twelve Thursday afternoons, I’ve had the chance not just to guide but to get to know a remarkable cohort of bachelor’s, master’s, and doctoral students.
Together, we tackled normative and descriptive questions raised by our expert lecturers, dialling in from across the European Union, the United Kingdom, and North America: Professor Ciaran Martin (University of Oxford), Professor Gabriella Coleman (Harvard University), Professor Andrew Martin (University of Oxford), Dr Monica Kello (King’s College London), Jan Króliński (interface), Anna Sophie Den Ouden (Virtual Routes), Dr James Shires (Virtual Routes), and Dr Max Smeets (Virtual Routes).
Reflecting on our weekly sessions from April to June, four observations stuck with me:
1. Knowing the technical and historical foundations will set you apart
Think of agentic AI systems (inter)acting autonomously with limited human supervision, the use of AI in drone warfare, or AI-orchestrated cyber espionage campaigns. It’s tempting to go straight for one of these shiny, timely topics. But the scope of our course gave our students ample time to dig into the basics of cybersecurity and AI.
James Shires (Virtual Routes) opened our very first session, ‘What is cybersecurity?’ with a detailed account of what we mean by ‘threat’ and ‘referent objects’, challenging students to engage with the underlying question of what we’re actually trying to secure, and from whom. While asking this may seem trivial, it is indispensable for a better understanding of, for instance, blurry boundaries between cyber conflict and cybercrime.
Andrew Martin (University of Oxford) dedicated a full session to systems security. Discussing why systems thinking matters in cybersecurity, he explained the basic principles of encryption and authentication, as well as the inherent, irreducible complexity that means cyber systems security will never be a solved problem.
Understanding these foundations will set you apart in addressing the shiny new cybersecurity or AI topic of the day later on. When it comes to explaining sophisticated attacks using machine learning models, for instance, analysis can be built on Anna Sophie Den Ouden’s (Virtual Routes) AI supply-chain deep dive, while Max Smeet’s (Virtual Routes) insights on how AI changes the cyber kill chain should be kept in mind when judging the prospects of vibe hacking.
My first observation? We need to do our groundwork properly, as the most interesting questions sit right on top of it.
2. The human factor matters!
We’ve learned that when securing the AI supply chain, for instance, cybersecurity can only be as strong as its weakest link. That link is often human. AI security is, in many ways, no different. The notion that cybersecurity is only as good as the individuals using the systems came up in almost every session in this course. Challenges often arise not from code but from the human factor: individuals and groups, their incentives, and underlying geopolitical power constellations.
Monica Kello (King’s College London) introduced us to these power constellations. Discussing the operational tendencies of different APTs, we covered decades of cyber conflict history in her class. Drawing on historical cases such as Volt Typhoon (2024), Shamoon (2012), and NotPetya (2017), she provided rich insight into the strategic thinking that shapes how NATO member states and actors such as China, North Korea, and Iran operate.
Zooming in on the groups behind cyber operations, Max Smeets (Virtual Routes) introduced us to his recent empirical research and conceptual thinking on what he calls the ‘ransomware trust paradox’ and the economics driving this unique ecosystem.
What stuck with me most was Gabriella Coleman’s (Harvard University) lecture on what some call ‘digital resistance movements’, and others refer to as ‘hacktivism’. Learning about her decades of in-depth anthropological research and continued face-to-face interaction with individuals involved with, for instance, Anonymous, powerfully underscored my second key observation: The human (security) factor matters profoundly in this often seemingly technical domain. Indeed, it matters more than we sometimes like to admit.
3. There’s no tidy answer to the role of AI
AI is a tool for defence as well as a weapon for offence. What I appreciated most in our discussions of AI was the mutual refusal of lecturers and students to offer a simple, decisive answer about the role of AI. Rest assured, AI is rapidly reshaping the cybersecurity landscape. And, as the majority of students voted in a poll during our final session with Ciaran Martin (University of Oxford), the biggest cybersecurity risks associated with AI may still lie ahead.
Asked which developments concern or excite them most when it comes to AI, students’ answers were both contrarian and thoughtful. Security implications of an AI race between authoritarian and democratic states, the use of AI for surveillance and information warfare, and increasingly powerful cyber attacks capable of disrupting critical national infrastructure were the most prominent concerns. Simultaneously, students were excited about future AI-driven cybersecurity solutions, AI-powered healthcare diagnostics, and applications in conflict resolution, agriculture or education.
My third observation: Especially regarding the role of AI, we benefit from honing controversial debates instead of settling them. Knowing the foundations of cybersecurity and AI significantly enriches our ability to do so.
4. The race to regulate is on, but where is the finish line?
Covering twelve different topics, our course exposed all students to a breadth of insight across the sometimes more technical, sometimes more political dimensions of cybersecurity and AI. One of my favourite sessions was in week 11, where we discussed how AI and cybersecurity are regulated.
I’m perfectly biased, of course, as this was also the session closest to my own doctoral research on economic statecraft and corporate influence in digital infrastructure governance. But this illustrates exactly how our course covers comprehensive foundations while simultaneously allowing students to zoom in on the niches they’re most fascinated by.
Making sense of the EU’s dense regulatory landscape, covering everything from the NIS/NIS 2 directives, the Cyber Resilience Act, the Cybersecurity Act, and the EU AI Act, we critically discussed whether over- or underregulating is the ‘bigger danger’, whether the guiding principle for AI regulation should be the type of application or the type of deployment, and which states were best at balancing safety with innovation.
Jan Króliński (interface/GovAI) explained the underlying incentives problem of cybersecurity regulation, which, as he argued, one can best approach through a microeconomic lens. This helped us grasp moral hazard, asymmetric information, and the negative externalities of botnets, for instance. Cautioning about how AI intersects with existing laws, rather than a legal vacuum, he suggested that, looking at frontier labs, for example, AI can also be interpreted as an incentives problem.
My fourth observation: Most, if not all, of these regulatory challenges are deeply normative. This is exactly where the interactive character of our course added the most value: by first equipping students with the foundational knowledge, and directly encouraging them to form an opinion and engage in critical discussion with peers and leading experts.
It’s about discussing not settling ideas
Our students are not supposed to leave with fixed ideas. Indeed, we intentionally dedicated our final session with Ciaran Martin (University of Oxford) to an interactive conversation across the many (many!) topics previously covered.
It was fascinating to see how our students left with perhaps even more, but significantly sharper questions and as part of a new multidisciplinary, international network of peers interested in engaging with similar challenges across silos.
