The method
Adversarial Creativity
Our method has a name: adversarial creativity — a translational adversarial inquiry in four movements, where the disciplined use of the attacker's mindset is both an investigative technique and an expressive medium.
Every Pwnshow investigation follows the same arc:
Conceive the process.
We design a novel adversarial process — an attack, an instrument, a market mechanism, an imaging technique — drawn from three decades of security engineering practice.
Match it to a researched subject.
The process is assessed for what it can reveal, then matched to a subject-matter we have researched in depth: a trust architecture, a regulatory gap, a social condition.
Execute to the standard of evidence.
Findings are produced to survive scrutiny — peer review where the claim is scientific, regulatory consultation where it is normative, exhibition where it is experiential.
Translate across registers.
Each investigation is rendered in the languages of its audiences: a paper, a standards contribution, a briefing, an image, an instrument. Translation is not an afterthought.
The result: knowledge meets conjecture, observation meets imagination, aisthesis meets poiesis.
Worked example — The Open-Weight Dilemma (INV-007)
A single Pwnshow investigation into the cyber risks of open-weight large language models produced: a research paper (arXiv:2505.17109), cited by the U.S. National Academies, Oxford Martin School AIGI, and UC Berkeley CLTC, among others; a policy article identifying the gaps in the EU AI Act and the draft GPAI Code of Practice; expert consultation to the EU AI Office that contributed to the substantial modification clause in the GPAI Code of Practice — protecting open-model providers from undue liability while making downstream actors responsible for their modifications; a talk at the ETSI Security Conference 2025 that led to an invitation to join ETSI TC SAI (Securing AI); a FOSDEM address carrying the findings to the open-source community; a talk to the European Pirate Party’s Think Twice conference on the governance of open-weight AI capabilities. One investigation; multiple registers; one demonstrable change in how Europe governs AI.