By Leonardo Bandarra.
On May 5, 2026, diplomats, researchers, and civil society representatives gathered in Conference Room A at UN Headquarters for the VeSPoTec-Hub side event “Verification and Knowledge in a Nuclear Weapons World”, supported by the German Ministry of Research, Technology, and Space (BMFTR).
The event was part of a broader VeSPoTec research network that brings together social and natural scientists to explore the role of and possibilities for verification in complex political environments.After a brief welcome and introduction to the VeSPoTec-Hub “Knowing in a Nuclear Weapons World”, we set out the core questions: How is nuclear knowledge produced, by whom, and with what power effects? How do secrecy, uncertainty, and risk shape decision-making? And what can civil society actually verify – and for whose benefit?

We framed the discussion around the idea that verification is not only a technical process of collecting data but also a deeply social and political practice. VeSPoTec’s wider research has emphasized that verification regimes depend on trust, perceptions of legitimacy, and broader institutional contexts, not just on the quality of sensors or algorithms.
Against the backdrop of eroding arms control regimes and new nuclear modernization programs, the panel examined how different communities – diplomats, scientists, and civil society – generate and interpret “nuclear knowledge,” and how this knowledge travels (or fails to travel) across institutional boundaries.

Secrecy and power

Dr.-Ing. Sara Al‑Sayed, from the Union of Concerned Scientists opened by grounding the conversation in the long history of civil society monitoring and societal verification. She revisited earlier proposals, such as those associated with Joseph Rotblat, to show that citizen-based observation and whistleblowing were envisioned as democratic safeguards well before today’s talk of open-source intelligence and satellite imagery.

From there, she moved to the politics of secrecy and power. Drawing on her work on AI, satellite remote sensing, and nuclear stability, Sara argued that data is not neutral: who controls sensors, funding, and interpretation often decides which risks are seen as “real” and which are dismissed. She highlighted cases where civil society data were questioned or sidelined, not because of lack of technical merit, but because they challenged entrenched power structures and existing strategic narratives. This raised a central theme for the rest of the event: the problem of trust – not just in states, but also in technologies, experts, and institutions.

Data, AI, and trust

Dr. Lyndon Burford then turned to the explosion of data and the promises and pitfalls of AI for verification. As Policy Fellow at BASIC and project manager of the Nuclear Transparency Inventory (NUTRI), he works on a new tool that systematically tracks how all nuclear-armed states disclose information about their arsenals and related policies. This experience informed his key message: more data does not automatically mean more confidence.
Lyndon unpacked three now-familiar AI challenges – interpretability, hallucinations, and alignment – and showed how they translate into the nuclear domain. AI systems can generate plausible but false patterns, hide their reasoning, and behave in ways that reflect opaque training data and objectives, raising serious questions about trust on AI-derived intelligence. He also discussed the problem of AI ‘integrity’ – the under-appreciated challenge of verifying that AI code has not been hacked or ‘self-improved’ by the AI itself. Lyndon outlined how cryptographic techniques and transparent, open algorithms might help build confidence in civil-society-generated information, for example by allowing independent verification of datasets, code, and analysis pipelines without exposing sensitive raw data.

Civil society’s role

Dr. Jana Baldus situated these technical and political issues in a broader institutional landscape. Drawing on her research on multilateral disarmament and nuclear (in)justice, she stressed that knowledge production around nuclear weapons is inherently political – shaped by funding flows, institutional mandates, and whose security is prioritized.
Jana highlighted the persistent imbalance between institutions in the Global North and Global South, both in terms of resources and in terms of whose perspectives define the research agenda. This affects what topics receive attention – for instance, whether nuclear-weapon-free zones and the lived experiences of affected communities are treated as central to nuclear order or as “side issues.” She warned of the danger that short funding cycles and market-driven priorities undermine creativity and long-term, critical research, precisely the kind of work needed to understand slow-burning risks and structural injustices. From this perspective, civil society’s role is not only to “add data” but also to question dominant frames, bring in marginalized perspectives, and keep long-term justice-oriented questions on the table when political debates narrow around immediate crises.

Connecting the dots

In the final intervention, Dr. Leonardo Bandarra tied the threads together under the theme of “Knowledge and Verification.” Building on the previous presentations, he explored how knowledge and non-knowledge are co-produced: what is left unknown, or rendered unknowable, can be as politically consequential as what is measured and reported.

He asked how trust and mistrust circulate between states, experts, and civil society, and how verification practices can either reinforce existing hierarchies or open space for more inclusive forms of nuclear governance. He argued for understanding verification as an ongoing, negotiated process that must grapple with uncertainty, conflicting narratives, and the risk of misreading emerging issues – including those related to new technologies and regional security arrangements such as nuclear-weapon-free zones.

Audience discussion

The Q&A session was unusually rich and practical. Participants raised questions about how far citizen verification can really go when critical data – for example, on enrichment activities or breakout times – remain under strict state control.
Others probed the role of commercial satellite companies and falling imagery prices in enabling “citizen verification.” The conversation quickly moved to the ethical and political implications: Who gets to interpret satellite data? How do we avoid crowd-sourced analysis reinforcing existing political narratives rather than challenging them? And what responsibilities do analysts have when their interpretations can influence public perceptions of crisis and compliance?

Throughout the event, one recurring thread was the need to resist short-termism. Verification challenges do not map neatly onto political cycles or funding calls, and neither do structural inequities in whose knowledge counts. For VeSPoTec-Hub and its partners, the side event underlined the importance of building sustained, interdisciplinary spaces where technical specialists, social scientists, diplomats, and civil society can work together on verification problems over time.