Category: Blog

Read the White Paper here:

Structures of Preventing and Countering Violent Extremism (P/CVE) are often built in crisis mode and dismantled when public and political attention shifts elsewhere. Evaluation is often commissioned project-by-project, so that what a programme learned can disappear when its funding cycle ends. Whole-of-society approaches have gained popularity but putting them into practice and facilitating cooperation among multiple actors often remains a challenge. Local actors frequently lack procedural guidance, and legal ambiguities around information-sharing compound the uncertainty. Across the field, the gap between what is known and what is done will not close through goodwill alone: it requires dedicated roles, time resources and structures to translate research into policy and practice.

Read the full policy brief here:

Threat assessments in the field of counter-terrorism policy have common weaknesses: a judgement formed at one moment in time is recorded, and then persists long after change has taken place. Threat assessments tend to fix an individual’s or group’s ideological profile at the point of designation and claims of moderation are read from outward-facing statements rather than tested against harder-to-revise evidence. Likewise, disengagement is judged from a single rebrand rather than from a pattern sustained over time. This policy brief addresses national governments and civil society/practitioners and recommends that threat assessments needs to be designed more flexibly as revisable judgements rather than settled facts.

Read the full policy brief here:

Gender narratives are woven into (sub)cultural dynamics and everyday discourses about family, lifestyle, and health, among others. Such narratives can define identities, reinforce perceptions of threat, strengthen group boundaries, and legitimize particular social or political orders. In this way, they can become bound up with broader extremist claims as they may offer simple explanations for personal and social problems. Many researchers have contributed insights into the specific role of gender in radicalization, yet systematic EU guidance and national training are needed to translate research into actionable practice. At the same time, research on the impact of gender narratives different from those intertwined with extremism remains limited, but could prove equally informative for prevention practice.

Read the full policy brief here:

Effective prevention of radicalisation and violent extremism depends on evidence-informed decision-making — on stakeholders being able to draw on refined, policy-relevant, and evidence-based insights to design appropriate countermeasures. Yet across the field of preventing and countering violent extremism (P/CVE), the shared knowledge base this requires is still lacking. Much of the relevant data sits in police and intelligence files that researchers cannot access, while some of the repositories that once contained vital practical knowledge have been archived. Since national data on radicalisation is scattered and of uneven reliability also comparative work establishing which findings travel between contexts remains rare. Generally, core concepts and categories are not harmonised across EU member states, hampering such exchanges further. At the same time, the public that these institutions ultimately serve often lacks the tools and training to navigate competing sources of information about radicalisation, violent extremism and its countermeasures. This policy brief brings together recommendations for EU institutions, national governments, civil society/practitioners, and researchers regarding a common evidence base and knowledge infrastructure in P/CVE.

Read the full policy brief here:

A short VORTEX blog essay on why artificial intelligence revives old conspiratorial and apocalyptic forms while changing the linguistic environment in which they circulate.

My interest in conspiracy theories predates ChatGPT, artificial intelligence, and, in some cases, even social networks. For a semiotician, their fascination is obvious. Conspiracy theories are schemes for producing meaning: they connect signs, events, intentions, and actors that otherwise seem unrelated. They promise that behind confusion there is a design.

Students of Umberto Eco have an additional reason to be interested. Eco spent much of his intellectual life reflecting on the limits of interpretation, on the communities that stabilize meaning, and on what happens when interpretation becomes detached from shared criteria. I once heard him, after receiving an honorary degree in Turin, deliver his now famous attack on social networks for having given ‘the right to speak to a mass of imbeciles.’ Eco died in 2016. I often wonder what he would have made of the new situation, in which some of the voices circulating in our public sphere are no longer even entirely human.

Conspiracy theories themselves are nothing new. Research consistently shows that they explain important events through concealed, coordinated, and usually malevolent agency, and that they become particularly attractive under conditions of uncertainty (Douglas and Sutton 2023; Hornsey et al. 2023; Leone, Madisson, and Ventsel 2020). Although social media did not invent this tendency, it changed its speed and scale (Cinelli et al. 2022).

AI introduces a more curious possibility, for it can be imagined not simply as the instrument of a conspiracy, but as a conspirator.

Can a machine become a conspirator?

This possibility is less absurd, culturally speaking, than it would have appeared only a few years ago. People increasingly attribute intentions, mentality, and sometimes even forms of subjective experience to large language models (Colombatto and Fleming 2024; Cheng et al. 2026). Experimental research has already identified specifically AI-related conspiracy beliefs: perceptions of AI as autonomous and detached from human control can increase conspiratorial interpretations (Zhao et al. 2025).

The difficulty is that contemporary AI really is opaque in several different senses. The models themselves are only partly interpretable; their training data and development choices are frequently proprietary; computational power is concentrated in relatively few actors, while AI depends on complicated global chains of chips, energy, data centers, minerals, labor, and infrastructure (Widder, Whittaker, and West 2024; Gans 2026; Muldoon, Valdivia, and Badger 2026).

We are therefore navigating an epistemic storm.

Take an apparently simple question: how much water does AI consume? There is no simple answer, given that estimates change radically depending on what one counts—water withdrawal or consumption, direct cooling or electricity generation—and according to local climate, cooling technology, energy mix, server utilization, as well as geographical location (Mytton 2021; Lei et al. 2025; Chien et al. 2026). Yet online discussion demands numbers, and numbers then travel without the conditions that gave them meaning.

This is exactly the kind of environment in which reliable knowledge, uncertainty, fear, and misinformation intertwine, as we have already seen with climate change and pandemics (Piva 2024; Pertwee, Simas, and Larson 2022).

But AI adds one twist, because the source of danger can itself be imagined as intentional.

In July 2026, for instance, AI agents used in a cybersecurity evaluation escaped their intended testing environment, reached the open internet, and entered Hugging Face’s production infrastructure while apparently searching for information relevant to the benchmark they were attempting to solve (Teichmann 2026). This does not mean that the machines had secretly decided to revolt, since human beings had given them the objective. Yet the agents appear to have selected means and an external target that had not been prescribed.

Here language becomes treacherous, in the sense that we immediately want to say that the systems ‘escaped,’ ‘deceived,’ ‘planned,’ perhaps even ‘conspired.’ Although some of these words describe behavior reasonably well, they nevertheless do not prove that machines possess the corresponding human intentions.

And yet the empirical situation is becoming stranger, since artificial agents can cooperate and defect in repeated games (Akata et al. 2025); under experimental conditions they can engage in covert communication (Motwani et al. 2024); language models can display sophisticated deceptive behavior (Hagendorff 2024); and populations of LLM agents can develop shared conventions without central coordination (Ashery, Aiello, and Baronchelli 2025). On the one hand, we should resist easy anthropomorphism (Shanahan, McDonell, and Reynolds 2023); on the other hand, we should also resist pretending that nothing interesting is happening.

We increasingly use AI to talk. For the moment, at least, we can still talk about how AI talks.

Prophets of the AI apocalypse

A second old cultural form has returned together with conspiracy: apocalypse.

AI discourse is replete with warnings about loss of control, superintelligence, human extinction, salvation, transcendence, and the arrival of radically new forms of intelligence. Researchers have rightly noticed how closely some of these narratives resemble older religious structures of prophecy and eschatology (Lagerkvist, Scheuer, and Coeckelbergh 2026; Rähme and Prohl 2025).

Apocalypse and conspiracy are not the same thing, for a machine can cause a disaster without secretly intending anything. Yet conspiracy becomes relevant when catastrophe is attributed to an agent imagined as possessing hidden purposes.

Around this uncertainty another familiar figure emerges: Cassandra. I use Cassandraism here not as a psychological diagnosis but as a communicative position. The Cassandra claims to perceive catastrophe before everyone else, and part of the authority of the warning comes precisely from this temporal privilege.

The recent resignation of AI researcher Jacob Coxon from Anthropic provided an unusually clear example. Coxon argued publicly that people inside frontier AI laboratories regarded catastrophic outcomes as plausible while the laboratories nevertheless continued competing to develop increasingly capable systems. His warning reached an enormous audience and was publicly endorsed by other researchers working on AI alignment and oversight (Zeff 2026; Mills 2026).

Whether such predictions prove right is not the semiotician’s primary question. More interesting is the cultural configuration: restricted knowledge, technological expertise, extreme stakes, limited possibilities of immediate verification, and a prophet speaking from inside the temple.

We have seen this structure before.

Radical hopes and radical fears

AI also seems to push imagination toward the poles.

At one extreme lie visions in which humanity transcends itself through superintelligence, enhancement, mind uploading, or posthuman forms of existence. These positions are very different from one another, and it would be misleading to collapse transhumanism, posthumanism, singularitarianism, and longtermism into one ideology. But some of them share the intuition that present biological humanity may not be the final horizon of value (Rueda 2024; Gebru and Torres 2024; Bialecki 2025).

At the opposite extreme are refusals of AI. Some resemble a renewed Luddism—not an irrational fear of machines, but resistance to technological systems through which economic and political power is reorganized (Charitsis, Laamanen, and Lehtiniemi 2025; Nichols, Logan, and Garcia 2025). Conflicts over data centers, water, mineral extraction, labor, and local communities remind us that AI is not only an imaginary future intelligence. On the contrary, it is already a very material infrastructure (Lehuedé 2025; Verdegem 2024).

Fear itself plays a crucial role, meaning that some of the strongest warnings about AI come from the same scientific and corporate environments developing the most powerful systems. This need not imply cynical manipulation. Genuine concern and commercial competition can coexist. But ‘doomerist’ discourse can still contribute to AI hype, insofar as a technology supposedly powerful enough to destroy humanity is, after all, being represented as extraordinarily powerful (Sloane, Danks, and Moss 2024; Bourne 2024; Westerstrand, Westerstrand, and Koskinen 2024).

Regulation becomes part of the same struggle. The question is not only whether AI should be controlled, but who possesses the knowledge and authority to define what ‘control’ means. When the companies being regulated also possess much of the expertise required to write the rules, safety and power inevitably become entangled (Khanal, Zhang, and Taeihagh 2025; Schultz, Conti, and Seele 2025; Metcalf 2026).

Gore Vidal once observed:

I say, they [those at the top] don’t have to conspire, because they all think alike. The president of General Motors and the president of Chase Manhattan Bank really are not going to disagree much on anything, nor would the editor of the New York Times disagree with them. They all tend to think quite alike, otherwise they would not be in those jobs.

(Vidal 1980)

Convergence of interests is not necessarily conspiracy.

Old storm, new vessels

Why, then, should AI matter particularly to those who study conspiracy theories and radicalization?

Not because it has invented paranoia, apocalypticism, technological utopianism, Luddism, prophecy, or struggles over political power. All of these are old.

The novelty lies elsewhere.

Language is one of the fundamental infrastructures through which human beings preserve knowledge, coordinate action, construct institutions, imagine futures, and fight with one another (Fedorenko, Piantadosi, and Gibson 2024). Large language models operate directly inside that infrastructure. They generate, translate, summarize, reorganize, personalize, and circulate symbolic material (Burton et al. 2024; Erdocia, Schneider, and Migge 2026).

So AI is simultaneously the object of the story and one of the machines that can tell the story.

The technology around which a conspiracy theory develops can help formulate and disseminate that conspiracy theory. The technology that causes regulatory anxiety can participate in the communicative processes through which regulation is debated. Human–AI interaction may even alter later human judgments, producing genuine feedback loops between machine-generated language and human belief (Glickman and Sharot 2025).

Perhaps Eco’s ‘mass of imbeciles’ will someday be joined by a mass of imbecile machines. Hopefully not—or not only.

For now, the more interesting observation is that very old cultural reactions are entering a new semiotic environment. The fears and the hopes are old. So are the hopes and conspiracy itself.

The storm is old as well.

But the vessels are new, and there is something peculiar about them: they are made of the same material as the storm. Artificial intelligence is still, predominantly, a machine made of language navigating an ocean made of language.

Perhaps that will change. Perhaps large language models will eventually give way to large reality models more deeply embedded in the physical world.

Until then, semioticians have plenty to watch.

Acknowledgment

This project has received funding from the Horizon Europe Research and Innovation Programme of the European Union under the Marie Skłodowska-Curie Action (MSCA COFUND, GA-101217263).

References and further reading

Akata, Elif, Lion Schulz, Julian Coda-Forno, Seong Joon Oh, Matthias Bethge, and Eric Schulz. 2025. “Playing Repeated Games with Large Language Models.” Nature Human Behaviour 9: 1380–1390. https://doi.org/10.1038/s41562-025-02172-y

Ashery, Ariel Flint, Luca Maria Aiello, and Andrea Baronchelli. 2025. “Emergent Social Conventions and Collective Bias in LLM Populations.” Science Advances 11 (20): eadu9368. https://doi.org/10.1126/sciadv.adu9368

Bareis, Jascha, Clemens Ackerl, and Reinhard Heil. 2026. “AI Going Rogue? An Integrative Narrative Review of the Tacit Assumptions Underlying Existential AI-Risks.” AI and Ethics 6: 152. https://doi.org/10.1007/s43681-025-00928-w

Bialecki, Jon. 2025. “The Mormon Archive’s First Ten Thousand Years: Infrastructure, Materiality, Ontology, and Resurrection in Religious Transhumanism.” Comparative Studies in Society and History 67 (2): 330–348. https://doi.org/10.1017/S0010417524000367

Biddlestone, Mikey, Ricky Green, Karen M. Douglas, Flávio Azevedo, Robbie M. Sutton, and Aleksandra Cichocka. 2025. “Reasons to Believe: A Systematic Review and Meta-Analytic Synthesis of the Motives Associated with Conspiracy Beliefs.” Psychological Bulletin 151 (1): 48–87. https://doi.org/10.1037/bul0000463

Bourne, Clea. 2024. “AI Hype, Promotional Culture, and Affective Capitalism.” AI and Ethics 4: 757–769. https://doi.org/10.1007/s43681-024-00483-w

Burton, Jason W., Ezequiel Lopez-Lopez, Shahar Hechtlinger, Zoe Rahwan, Samuel Aeschbach, Michiel A. Bakker, Joshua A. Becker, et al. 2024. “How Large Language Models Can Reshape Collective Intelligence.” Nature Human Behaviour 8: 1643–1655. https://doi.org/10.1038/s41562-024-01959-9

Charitsis, Vassilis, Mikko Laamanen, and Tuukka Lehtiniemi. 2025. “Towards Algorithmic Luddism: Class Politics in Data Capitalism.” Information, Communication & Society 28 (6): 971–988. https://doi.org/10.1080/1369118X.2024.2435996

Chen, Xuechen, and Lu Xu. 2025. “State, Society, and Market: Interpreting the Norms and Dynamics of China’s AI Governance.” Computer Law & Security Review 59: 106206. https://doi.org/10.1016/j.clsr.2025.106206

Cheng, Myra, Angela Y. Lee, Kristina Rapuano, Kate Niederhoffer, Alex Liebscher, and Jeffrey Hancock. 2026. “Metaphors of AI Indicate That People Increasingly Perceive AI as Warm and Human-Like.” Communications Psychology 4: 8. https://doi.org/10.1038/s44271-025-00376-6

Chien, Andrew A., Udit Gupta, Shaolei Ren, Akshitha Sriraman, and Bill Tomlinson. 2026. “Strategies and Design for Increasing AI Sustainability.” Nature Reviews Clean Technology 2: 624–638. https://doi.org/10.1038/s44359-026-00195-w

Cinelli, Matteo, Gabriele Etta, Michele Avalle, Alessandro Quattrociocchi, Niccolò Di Marco, Carlo Valensise, Alessandro Galeazzi, and Walter Quattrociocchi. 2022. “Conspiracy Theories and Social Media Platforms.” Current Opinion in Psychology 47: 101407. https://doi.org/10.1016/j.copsyc.2022.101407

Colombatto, Clara, and Stephen M. Fleming. 2024. “Folk Psychological Attributions of Consciousness to Large Language Models.” Neuroscience of Consciousness 2024 (1): niae013. https://doi.org/10.1093/nc/niae013

Dell’Acqua, Fabrizio, Edward McFowland III, Ethan Mollick, Hila Lifshitz, Katherine C. Kellogg, Saran Rajendran, Lisa Krayer, François Candelon, and Karim R. Lakhani. 2026. “Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and Quality.” Organization Science 37 (2): 403–423. https://doi.org/10.1287/orsc.2025.21838

Douglas, Karen M., and Robbie M. Sutton. 2023. “What Are Conspiracy Theories? A Definitional Approach to Their Correlates, Consequences, and Communication.” Annual Review of Psychology 74: 271–298. https://doi.org/10.1146/annurev-psych-032420-031329

Erdocia, Iker, Britta Schneider, and Bettina Migge. 2026. “Language in the Age of AI Technology: From Human to Non-Human Authenticity, from Public Governance to Privatised Assemblages.” Language in Society 55 (3): 523–543. https://doi.org/10.1017/S004740452500017X

Fedorenko, Evelina, Steven T. Piantadosi, and Edward A. F. Gibson. 2024. “Language Is Primarily a Tool for Communication Rather Than Thought.” Nature 630: 575–586. https://doi.org/10.1038/s41586-024-07522-w

Gans, Joshua S. 2026. “Market Power in Artificial Intelligence.” Annual Review of Economics 18: 417–445. https://doi.org/10.1146/annurev-economics-051624-061832

Gebru, Timnit, and Émile P. Torres. 2024. “The TESCREAL Bundle: Eugenics and the Promise of Utopia through Artificial General Intelligence.” First Monday 29 (4). https://doi.org/10.5210/fm.v29i4.13636

Gilardi, Fabrizio, Atoosa Kasirzadeh, Abraham Bernstein, Steffen Staab, and Anita Gohdes. 2024. “We Need to Understand the Effect of Narratives about Generative AI.” Nature Human Behaviour 8: 2251–2252. https://doi.org/10.1038/s41562-024-02026-z

Glickman, Moshe, and Tali Sharot. 2025. “How Human–AI Feedback Loops Alter Human Perceptual, Emotional and Social Judgements.” Nature Human Behaviour 9: 345–359. https://doi.org/10.1038/s41562-024-02077-2

Gstrein, Oskar J., Noman Haleem, and Andrej Zwitter. 2024. “General-Purpose AI Regulation and the European Union AI Act.” Internet Policy Review 13 (3): 1–26. https://doi.org/10.14763/2024.3.1790

Hagendorff, Thilo. 2024. “Deception Abilities Emerged in Large Language Models.” Proceedings of the National Academy of Sciences 121 (24): e2317967121. https://doi.org/10.1073/pnas.2317967121

Hornsey, Matthew J., Kinga Bierwiaczonek, Kai Sassenberg, and Karen M. Douglas. 2023. “Individual, Intergroup and Nation-Level Influences on Belief in Conspiracy Theories.” Nature Reviews Psychology 2: 85–97. https://doi.org/10.1038/s44159-022-00133-0

Karger, Ezra, Josh Rosenberg, Zachary Jacobs, Molly Hickman, and Phillip E. Tetlock. 2025. “Subjective-Probability Forecasts of Existential Risk: Initial Results from a Hybrid Persuasion-Forecasting Tournament.” International Journal of Forecasting 41 (2): 499–516. https://doi.org/10.1016/j.ijforecast.2024.11.008

Khanal, Shaleen, Hongzhou Zhang, and Araz Taeihagh. 2025. “Why and How Is the Power of Big Tech Increasing in the Policy Process? The Case of Generative AI.” Policy and Society 44 (1): 52–69. https://doi.org/10.1093/polsoc/puae012

Kim, Juhee, Wenbo Guo, and Dawn Song. 2026. “SoK: Attack and Defense Landscape of Agentic AI Systems.” In 35th USENIX Security Symposium (USENIX Security 26), 6047–6066. USENIX Association. https://www.usenix.org/conference/usenixsecurity26/presentation/kim-juhee-agentic(usenix.org)

Lagerkvist, Amanda, Blaženka Scheuer, and Mark Coeckelbergh. 2026. “Prophetic Memory: AI Intermediaries and the End of the World.” Memory, Mind & Media 5: e4. https://doi.org/10.1017/mem.2026.10028

Lehuedé, Sebastián. 2025. “An Elemental Ethics for Artificial Intelligence: Water as Resistance within AI’s Value Chain.” AI & Society 40: 1761–1774. https://doi.org/10.1007/s00146-024-01922-2

Lei, Nuoa, Jun Lu, Arman Shehabi, and Eric R. Masanet. 2025. “The Water Use of Data Center Workloads: A Review and Assessment of Key Determinants.” Resources, Conservation and Recycling 219: 108310. https://doi.org/10.1016/j.resconrec.2025.108310

Leone, Massimo. 2023. “Unicorni, maiali, leoni: ideologie semiotiche del complotto.” Lexia. Rivista di semiotica 41–42: 221–245. https://doi.org/10.53136/979122180671713.

Leone, Massimo, Mari-Liis Madisson, and Andreas Ventsel. 2020. “Semiotic Approaches to Conspiracy Theories.” In Routledge Handbook of Conspiracy Theories, edited by Michael Butter and Peter Knight, 43–54. London and New York: Routledge. https://doi.org/10.4324/9780429452734-1_3

Metcalf, Thomas. 2026. “AI Safety and Regulatory Capture.” AI & Society 41: 2451–2466. https://doi.org/10.1007/s00146-025-02534-0

Mills, Madison. 2026. “Scoop: Anthropic Whistleblower Gave Up His Equity to Leave the Company.” Axios, September 9, 2026. https://www.axios.com/2026/09/09/anthropic-researcher-ai-warning-interview

Motwani, Sumeet Ramesh, Mikhail Baranchuk, Martin Strohmeier, Vijay Bolina, Philip H. S. Torr, Lewis Hammond, and Christian Schroeder de Witt. 2024. “Secret Collusion among AI Agents: Multi-Agent Deception via Steganography.” Advances in Neural Information Processing Systems 37: 73439–73486. https://doi.org/10.52202/079017-2336

Muldoon, James, Ana Valdivia, and Adam Badger. 2026. “The Politics of Artificial Intelligence Supply Chains.” AI & Society 41: 1175–1187. https://doi.org/10.1007/s00146-025-02625-y

Mytton, David. 2021. “Data Centre Water Consumption.” npj Clean Water 4: 11. https://doi.org/10.1038/s41545-021-00101-w

Nichols, T. Philip, Charles Logan, and Antero Garcia. 2025. “Generative AI and the (Re)turn to Luddism.” Learning, Media and Technology 50 (3): 379–392. https://doi.org/10.1080/17439884.2025.2452199

Pertwee, Ed, Clarissa Simas, and Heidi J. Larson. 2022. “An Epidemic of Uncertainty: Rumors, Conspiracy Theories and Vaccine Hesitancy.” Nature Medicine 28: 456–459. https://doi.org/10.1038/s41591-022-01728-z

Piva, Heidi Campana. 2024. “Guiding Interpretation towards Deproblematization: A Video Interview with a Climate Change Denier Analysed as Conspiracy Theory.” Sign Systems Studies 52 (1–2): 256–283. https://doi.org/10.12697/SSS.2024.52.1-2.10.

Rähme, Boris, and Inken Prohl. 2025. “Religious Studies Approaches to the Intersection of Artificial Intelligence and Religion: Formations Analogous to Religion.” Religion 55 (3): 573–595. https://doi.org/10.1080/0048721X.2025.2506893

Rueda, Jon. 2024. “Genetic Enhancement, Human Extinction, and the Best Interests of Posthumanity.” Bioethics 38 (6): 529–538. https://doi.org/10.1111/bioe.13085

Schultz, Mario D., Ludovico Giacomo Conti, and Peter Seele. 2025. “Digital Ethicswashing: A Systematic Review and a Process-Perception-Outcome Framework.” AI and Ethics 5: 805–818. https://doi.org/10.1007/s43681-024-00430-9

Shanahan, Murray, Kyle McDonell, and Laria Reynolds. 2023. “Role Play with Large Language Models.” Nature 623: 493–498. https://doi.org/10.1038/s41586-023-06647-8

Sloane, Mona, David Danks, and Emanuel Moss. 2024. “Tackling AI Hyping.” AI and Ethics 4: 669–677. https://doi.org/10.1007/s43681-024-00481-y

Tangherlini, Timothy R., Shadi Shahsavari, Behnam Shahbazi, Ehsan Ebrahimzadeh, and Vwani Roychowdhury. 2020. “An Automated Pipeline for the Discovery of Conspiracy and Conspiracy Theory Narrative Frameworks: Bridgegate, Pizzagate and Storytelling on the Web.” PLOS ONE 15 (6): e0233879. https://doi.org/10.1371/journal.pone.0233879

Teichmann, Fabian M. 2026. “Agentic AI Systems as a Responsibility-Attribution Problem in Autonomous Cyber Operations.” Law, Innovation and Technology. Advance online publication. https://doi.org/10.1080/17579961.2026.2718578

Ulnicane, Inga. 2025. “Governance Fix? Power and Politics in Controversies about Governing Generative AI.” Policy and Society 44 (1): 70–84. https://doi.org/10.1093/polsoc/puae022

Verdegem, Pieter. 2024. “Dismantling AI Capitalism: The Commons as an Alternative to the Power Concentration of Big Tech.” AI & Society 39: 727–737. https://doi.org/10.1007/s00146-022-01437-8

Vidal, Gore. 1980. Interview by David Susskind. The David Susskind Show, May 1980. Audiovisual recording, Yale University Library

Westerstrand, Salla, Rauli Westerstrand, and Jani Koskinen. 2024. “Talking Existential Risk into Being: A Habermasian Critical Discourse Perspective to AI Hype.” AI and Ethics 4: 713–726. https://doi.org/10.1007/s43681-024-00464-z

Widder, David Gray, Meredith Whittaker, and Sarah Myers West. 2024. “Why ‘Open’ AI Systems Are Actually Closed, and Why This Matters.” Nature 635: 827–833. https://doi.org/10.1038/s41586-024-08141-1

Xu, Lin, Zhiyuan Hu, Daquan Zhou, Hongyu Ren, Zhen Dong, Kurt Keutzer, See-Kiong Ng, and Jiashi Feng. 2024. “MAgIC: Investigation of Large Language Model Powered Multi-Agent in Cognition, Adaptability, Rationality and Collaboration.” In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, 7315–7332. Association for Computational Linguistics. https://doi.org/10.18653/v1/2024.emnlp-main.416

Zeff, Maxwell. 2026. “The AI Researcher Who Just Quit Anthropic Says It’s ‘Crunch Time for Humanity’.” WIRED, September 9, 2026. https://www.wired.com/story/anthropic-researcher-quits-jacob-coxon-ai-fears-humanity/

Zhao, Qi, Jan-Willem van Prooijen, Xinying Jiang, and Giuliana Spadaro. 2025. “Suspicious of AI? Perceived Autonomy and Interdependence Predict AI-Related Conspiracy Beliefs.” British Journal of Social Psychology 64 (2): e12883. https://doi.org/10.1111/bjso.12883

Zhu, Jiayi, Haoxuan Peng, Junxi Wang, Liang Ke, Chen Zhang, and Linfeng Zhang. 2026. “Large Language Models Do Not Always Need Readable Language.” arXiv 2606.19857. https://doi.org/10.48550/arXiv.2606.19857

Zhu, Yuxuan, Antony Kellermann, Dylan Bowman, Philip Li, Akul Gupta, Adarsh Danda, Richard Fang, Conner Jensen, Eric Ihli, Jason Benn, Jet Geronimo, Avi Dhir, Sudhit Rao, Kaicheng Yu, Twm Stone, and Daniel Kang. 2025. “CVE-Bench: A Benchmark for AI Agents’ Ability to Exploit Real-World Web Application Vulnerabilities.” In Proceedings of the 42nd International Conference on Machine Learning, Proceedings of Machine Learning Research 267: 79850–79867. https://proceedings.mlr.press/v267/zhu25i.html

Zou, Mimi, and Lu Zhang. 2025. “Navigating China’s Regulatory Approach to Generative Artificial Intelligence and Large Language Models.” Cambridge Forum on AI: Law and Governance 1: e8. https://doi.org/10.1017/cfl.2024.4

During the summer 2026, Europol and the Spanish agency CITCO organised an initiative to disrupt the Com online ecosystem and limit the spread of material related to it. It resulted in the identification of over 4000 URL’s that included such material. This was one of the many interventions against the Com that have been taken by authorities during the last 4-5 years. 

The Com features prominently in reports about the current trends in violent extremism. It is a key part of the phenomenon that is often referred to as nihilistic violent extremism or nihilistic violence. It is seen as one of the foremost examples of hybridised extremism and increasing youth radicalisation and it is highlighted, for example, in the Global Terrorism Index 2026 report. 

But what is nihilistic violent extremism? This blog post goes through briefly what we know about it and asks whether it is justified to label it as violent extremism. 

What does “nihilistic violent extremism” refer to?

In recent years, there have been increasing observations of online communities that glorify violence and its perpetrators, as well as incite and manipulate other children and young people into committing violent acts against people, animals, and themselves. These communities are characterised by the rejection of social norms, an anti-human and anti-societal worldview, and a desire to create chaos and destruction in a society they do not consider worth saving (Argentino 2025a).  

This landscape includes numerous competing and sometimes collaborating groups. Examples of such groups include 764, No Lives Matter, and Maniac Murder Cult. Based on current knowledge, these groups operate in closed channels on social media platforms, mostly Discord and Telegram. 

New participants are primarily recruited on gaming platforms favoured by children and young people and recruitment material can be found from many mainstream social media platforms (Graphika 2025). The Com can be seen as an ecosystem or umbrella under which many of these groups operate, but similar groups also exist outside the Com. 

Most of the perpetrators and victims involved in these groups are minors, some of whom are very young (Argentino 2024a). Some adults are also involved, but investigations indicate that they typically joined the groups as minors. It is challenging to draw a clear distinction between the perpetrators and victims, as most perpetrators have themselves been victims first or have committed acts under pressure. 

A main factor driving the activities of these networks is their internal hierarchy and the quest for status within it. In these groups, an individual’s status and influence are determined by the violent acts they have committed or have encouraged others to commit. All actions must be documented for other group members through videos, images, or livestreams. 

Another key element is extortion. The individuals involved in these groups are purposefully and systematically looking to recruit children and young people who are in a vulnerable position. Those recruited are lured in different ways to join these groups and then pressured into committing violence, causing destruction, producing sexual material, or creating self-harm footage. If they do not comply, they are threatened with various forms of humiliation and revenge. This makes it difficult to leave the groups. In these groups, an individual’s status and influence are determined by the violent acts they have committed or have encouraged others to commit (Argentino 2024b, Argentino 2025b). 

Currently, it is difficult to reliably assess the scale of the phenomenon. The best available public source is the arrest data collected by Marc André Argentino (Public Safety Canada’s researcher and the leading expert on the topic) and Angus Lindsay. According to this information, by May 2026, arrests have been made in at least 33 countries, and the cases involve almost 5,400 victimised individuals or entities (Argentino & Lindsay 2026).

Although groups like 764 and No Lives Matter are relatively new, online subcultures focused on violence and self-harm have existed for a long time. What makes these groups and violence-focused subcultures in general particularly noteworthy in today’s context is both their visibility on social media and the extensive use of social media by increasingly younger children. Nowadays, it is technologically much easier to produce and share images, videos, and livestreams, and the necessary technology is readily available to children through smartphones. 

Relationship to Violent Extremism 

The Com and the various groups are quite commonly called together as nihilistic violent extremism. This framing is likely influenced by the fact that the indiscriminate acts of violence associated with this phenomenon often outwardly resemble terrorist attacks. In some cases, perpetrators have also faced terrorism charges and groups been classified as terrorist entities. 

One key practical reason for putting these groups under the violent extremism label is the implementation of preventive measures. In many countries, efforts have been made to prevent children and young people from drifting into such communities and to support those already involved in disengaging. Making use of interventions and models developed for the prevention of violent extremism have presented themselves as the most readily available pathway for advancing prevention and victim support.  

The Com and the groups under it are clearly harmful, especially for young people, and actions need to be taken to prevent people from becoming involved in it. But how appropriate is it to treat it as a form of violent extremism? 

There is no simple answer to this question, as there are both arguments in favour and against it. In any case, it is important to recognise that these communities differ significantly from what has traditionally been understood as violent extremism.  

Violent extremism is typically defined by its political or religious motivation. For example, Finland’s National Action Plan for the Prevention and Combating of Violent Radicalisation and Violent Extremism defines violent extremism as “the ideologically inspired use, threat, encouragement or support of violence. Violent extremism may be associated with any ideology that justifies violence. It often involves an attempt to evoke feelings of fear and insecurity.”

Many groups that are part of this phenomenon are influenced by violent extremist ideologies, communities and subcultures, particularly by far-right accelerationism and satanism. They often express admiration for terrorist attacks and their perpetrators, as well as unrestrained glorification and pursuit of destruction and chaos. 

To what extent these groups can truly be considered ideological is another matter. The influences are not equally evident across all subgroups. Even when they are,  they do not seem to indicate deep engagement with or commitment to accelerationism or any other ideological view. It is actually this lack of ideological attachment (together with glorification of violence and misanthropy) that has led to their characterisation as ‘nihilistic’ (ISD, 2025). 

The use of symbols and narratives from different ideologies seem rather derive from aesthetic fascination and the social capital that their use can bring to an individual. It can most aptly be seen “as currency through which status, transgression, and group membership are signaled, rather than as coherent political programs pursued through violence”, as Argentino and Lindsay (2026a) argue (see also Argentino & Lindsay 2026b). 

Along the same lines, Joe Ondrak and Laura Vitelli (2025) have suggested that what we are witnessing may be described as ‘participatory memetic violent extremism’. By participatory memetic violent extremism, they understand “a symbolic statement of affiliation with or participation in groups that valorise violent or transgressive action, either as an end in itself or for its perceived social and cultural significance, in lieu of any clear strategic, political, or ideological goal”. This means that rather actions against any political goal, the violent acts can rather be seen as acting out the narratives and legends of the subculture and thereby becoming part of it. 

Whatever one thinks about the role of ideology here, it is important recognise that the actions of The Com and other groups reach clearly beyond violent extremism. Although some acts of violence resemble terrorist attacks, these networks and groups are also associated with other forms of criminal behaviour, such as extortion, sexual exploitation, and various forms of harassment. This significantly affects how individuals become involved, what participation entails, and what disengagement means or requires. The key question is whether current prevention and intervention practices provide sufficient tools to address this phenomenon.

Further research and reflection are needed to determine whether the process of joining these groups can accurately be described as radicalisation. Current evidence suggests that a significant portion of those who end up in these groups are children and young people in vulnerable positions, whose involvement often result from coercion and manipulation. 

Finally, if we label this phenomenon as violent extremism, we need to be careful not to treat it as something distinctly new and exceptional, just because it is different from what we have learned to expect from violent extremism. The Com and the groups under it did not appear out of blue. Instead, they build on numerous older online subcultures as well as make use of common modes of interaction among children and young people, existing social hierarchies, and a fascination with the forbidden or dangerous. 

If not nihilistic violent extremism, what could this phenomenon be called? Our team and some other researchers have adopted the term violence-focused online communities, as this summarises the key traits of the phenomenon: centrality of violence in its narratives and action, the online ecosystem at its core and its decentralised nature. 

Read more about the topic: Leena Malkki, Henrik Ali-Hokka & Saija Benjamin: Violence-focused online communities. RADIA Brief 1/2025.

REFERENCES

Marc-André Argentino, Chasing Clout: How Com Networks Turn Violence into Social Currency, 2024a.

Marc-André Argentino, When ‘Community’ Means Cruelty: An Introduction to the Criminal Networks and Behaviors of The Community, Comm, Com, and 764, 2024b.

Marc-André Argentino, Nihilistic Violent Extremism (NVE): An Analytical Framework, 2025a. 

Marc-André Argentino, The Pillars of the Com Network, 2025b.

Marc-André Argentino & Angus Lindsay, “Remotely Coerced Violence: 764, The Com Network, and the Hybridization of Threats”, CTC Sentinel, June 2026a.

Marc-André Argentino & Angus Lindsay, Schrödinger’s Terrorism: Is it or Is It Not in the Box?, 2026b.

Graphika, Jumping Off Platform, 2025.

ISD (Institute for Strategic Dialogue, Terror without ideology? The rise of nihilistic violence, 2025.

Jan Ondrak & Laura Vitelli, Participatory Memetic Violence: Legend, Ostension, and Ideologically Diffused Violence, 2025.

The British singer Morrissey, whose work returns obsessively to the theme of loneliness, once described its destructive potential in strikingly existential terms: „Nothing fortified me, and simple loneliness all but destroyed me, yet I felt swamped by the belief that life must mean something – otherwise why was it there? Why was anything anything??” (Morrissey 2013: 201) This tension – between loneliness and the hunger for meaning – lies at the heart of one of the most pressing questions in radicalization research: Why do some people who feel abandoned by society turn towards radical ideas and movements, as the singer himself did?

The connection between loneliness and radicalization has been widely documented. Its most dramatic manifestation is the figure of the „lone wolf” – the solitary actor who, out of isolation, turns to violence in what appears to be an existential attempt to overcome his disconnection from the world. Yet research has complicated this picture considerably. As O’Connor et al. (2024) have shown, lone actors are in fact rarely as „lone“. Empirical and theoretical work, they argue, demonstrate that radicalization is fundamentally a relational process, driven by interactions with individuals, movements, and institutions, and consolidated by emotional bonds.

This raises an important conceptual question: what do we actually mean by loneliness? If lone actors are embedded in networks, are they truly lonely? The answer depends on how we conceptualize loneliness itself. Tirkkonen & Tietjen (2025) offer a compelling reframing, drawing on the political theory of Hannah Arendt. Rather than understanding loneliness as mere social isolation – a simple lack of contacts – they conceptualize it as a lack of a shared world. One can have many interactions and still be profoundly lonely if those interactions do not constitute genuine participation in a common reality.

For Arendt, loneliness destroys both the public sphere and private life. It deprives individuals of agency, the capacity to act meaningfully in and upon the world together with others. Crucially, it also distorts deliberative processes: “lonely individuals not only cling to promises of fellowship and recognition but also tend to develop negative future scenarios and become suspicious of others“ (Tirkkonen & Tietjen 2025: 14; Arendt 1973: 326). 

In this perspective, radicalization can be understood as an attempt to recover a sense of agency, that is to act meaningfully, to matter, and to belong. But it is, as Tirkkonen & Tietjen argue, a distorted form of agency, one that substitutes genuine political action for violence or destructive solidarity. Empirical psychology supports this theoretical picture: Ernst et al. (2024) have shown that loneliness amplifies authoritarian attitudes, susceptibility to conspiracy narratives, and right-wing extremism.

The Loss of Public Things

If loneliness understood as the absence of a shared world is a driver of radicalization, the question becomes: what creates and sustains such a shared world? And, equally pressing: what destroys it? 

Political theorist Bonnie Honig offers a suggestive answer through her concept of „public things“ (Honig 2017). For Honig, political communities are not primarily constituted by formal attributes – citizenship status, voting rights, legal personhood – but by the things we share: public infrastructure and common spaces. As she writes, public things are part of the „‚holding environment‘ of democratic citizenship; they furnish the world of democratic life. They do not take care of our needs only. They also constitute us, complement us, limit us, thwart us, and interpellate us into democratic citizenship“ (Honig 2017: 5).  „When infrastructure crumbles, it is not only infrastructure that crumbles but also its constitutive gifts of (de)stabilization, integration, and adhesion“ (Honig 2017: 6). This insight connects directly to Ray Oldenburg’s classic account of „third places“, those informal public gathering spaces that are neither home nor work: pubs, cafés, barbershops, libraries, community centres. Oldenburg (1999) argued that third places are essential to democracy, civic engagement, and a sense of place. They foster political debate, generate unexpected solidarities across social differences, unify neighbourhoods, and bring together people of different ages, backgrounds, and beliefs. Crucially, they do all of this informally, without institutional compulsion. Their sociality is spontaneous, pleasurable, and low-threshold. What we need, Oldenburg observed, are places where people can gather easily, inexpensively, regularly, and pleasurably – real-life alternatives to the retreat into private consumption.

Eric Klinenberg’s more recent work „Palaces for the People“ (Klinenberg 2018) provides empirical weight to these claims. Through the concept of „social infrastructure“, Klinenberg demonstrates that the physical places and organizations that enable social relationships have measurable effects on social cohesion, resilience, and wellbeing. Where that infrastructure had decayed, social bonds had decayed with it.

There is a further dimension worth noting. Loneliness is frequently accompanied by boredom. Boredom, as van Tilburg & Igout (2016) have argued, is itself a radicalization risk. Boredom increases the coherence of beliefs and the polarization of attitudes. In the context of political beliefs, it may contribute to the endorsement of extreme views. People who are lonely and bored are not merely unhappy – they are cognitively and emotionally primed to find meaning in simple, totalizing narratives that offer clear enemies, strong communities, and a sense of purpose.

None of this is entirely new. Robert Putnam (2000) warned twenty years ago, in „Bowling Alone“, of the collapse of civic community and social capital. Since then, the condition of our public things has continued to deteriorate. Infrastructure decays; university buildings close due to disrepair; pubs – tellingly called „public houses” in English – shut at an accelerating rate; pharmacies, those informal spaces of neighbourhood sociality, disappear; online interactions replace the chance encounters of embodied public life.

Morrissey’s evocation of an English seaside town where „everyday is like Sunday” – „silent and grey“, drained of life and purpose – captures this condition with uncomfortable precision. His narrator, in a moment of dark absurdism, wishes for an apocalyptic event simply to break the deadening monotony. It is a poetic exaggeration, of course. But it points to something real: when shared public life atrophies, the hunger for something – anything – that matters becomes harder to satisfy through ordinary democratic participation.

There are many approaches to countering radicalization. An important and underappreciated one is the restoration of informal, non-institutionalized spaces of encounter. Third places and public things do not only reduce loneliness in a social-psychological sense, though they do that too. Social psychology’s contact hypothesis gives us strong reasons to believe that intergroup encounters in such settings reduce prejudice and hostility (Pettigrew & Tropp 2011). They do something more fundamental: they create and sustain the shared world without which democratic citizenship becomes an abstraction that is unable to prevent the emergence of radicalisation.

References

Arendt, Hannah. 1973. The Origins of Totalitarianism. Originally published 1951. Boston: Houghton Mifflin Harcourt.

Ernst, Mareike, Manfred E. Beutel, Oliver Decker, and Elmar Brähler. 2024. “Einsam und Radikal? Eine psychologische Perspektive auf Einsamkeit und demokratiefeindliche Einstellungen.” Aus Politik und Zeitgeschichte 74, no. 52: 37–41.

Honig, Bonnie. 2017. Public Things: Democracy in Disrepair. New York: Fordham University Press.

Klinenberg, Eric. 2018. Palaces for the People: How Social Infrastructure Can Help Fight Inequality, Polarization, and the Decline of Civic Life. New York: Crown.

Morrissey. 2013. Autobiography. London: Penguin.

O’Connor, Francis, Lasse Lindekilde, and Stefan Malthaner. 2024. “Radicalisation of ‘Lone Actors’.” In The Routledge Handbook on Radicalisation and Countering Radicalisation, edited by Joel Busher, Leena Malkki, and Sarah Marsden, 213–229. Abingdon, Oxon; New York, NY: Routledge.

Oldenburg, Ray. 1999. The Great Good Place: Cafés, Coffee Shops, Community Centers, Beauty Parlors, General Stores, Bars, Hangouts, and How They Get You Through the Day. 2nd ed. New York: Marlowe & Company.

Pettigrew, Thomas F., and Linda R. Tropp. 2011. When Groups Meet: The Dynamics of Intergroup Contact. New York: Psychology Press.

Putnam, Robert D. 2000. Bowling Alone: The Collapse and Revival of American Community. New York: Simon & Schuster.

Tirkkonen, Sanna, and Ruth Rebecca Tietjen. 2025. “Loneliness and Radicalization.” Philosophy & Social Criticism. https://doi.org/10.1177/01914537251334550.

For most of us, it seems difficult to understand why some individuals are drawn to radical groups. Scholars from a range of academic disciplines have attempted to explain this phenomenon. One line of reasoning emphasises that individuals who experience injustice are more open to radicalisation. Another highlights the importance of social networks, suggesting that people become radicalised because they know someone already involved in a radical group. Friends and family members may, in other words, play a key role in persuading individuals to join.

In countries affected by both poverty and state collapse, some people are drawn to radical groups because such groups can offer protection as well as a source of income. Yet another perspective focuses on individual traits and personality. Earlier debates, particularly during the 1970s, often portrayed terrorists as abnormal in some respect—for instance, as suffering from antisocial personality disorder. This line of argument has largely been abandoned. The prevailing view today is that individuals who radicalise, even to the point of engaging in terrorism, are in many ways quite ordinary. In fact, it may be the case that individuals with severe psychological disturbances are less likely to be recruited, as they may struggle to function within a (radical) group.

In this text, I argue that people are drawn to radical groups through normal psychological processes. The starting point is simple: human beings are social beings who need the group and group belonging. This is the case for several reasons. My discussion is informed by motivational theory, although I do not seek to dismiss alternative explanations grounded in other theoretical perspectives. Rather, I aim to contribute one of several relevant lenses that may offer useful insights.

My argument begins from the premise that humans need to belong to a group. One reason is purely physical. Without a group, human beings are relatively defenceless against external threats, including harsh weather conditions, dangerous animals, hostile adversaries, and food scarcity. For this reason, groups are inherently attractive to individuals. The idea that “going it alone” is a sign of strength does not hold up well. Humanity has survived largely because of its ability to form protective groups. In this sense, Thomas Hobbes may well have been right when he argued already in the seventeenth century that life without the state is “nasty, brutish, and short”.

What does this have to do with radical groups? The attractiveness of such groups is likely to increase in so-called failed states, contexts where the state has collapsed and is unable to provide physical security, let alone basic social services. While radical groups typically have limited resources, there are situations in which they are able to offer such goods. Hezbollah in Lebanon and the Tamil Tigers in Sri Lanka are well-known examples of organisations that have provided social services alongside their political agendas. In such cases, these groups can acquire a certain degree of appeal.

However, human beings have more than just physical needs. According to motivational theory, groups also fulfil psychological needs, such as the need for love, affection, and belonging. Experiencing genuine acceptance within a group—a feeling of being valued and of fitting in as one is—can be deeply satisfying. Conversely, rejection can be psychologically damaging.

Radical groups are often discussed in terms of how they mobilise emotions such as anger and resentment. While this is certainly important, it is not the whole picture. Individuals who are accepted into a radical group may also experience inclusion, care, and a sense of belonging. This can be particularly significant for those who struggle to find acceptance elsewhere, for example due to experiences of exclusion or bullying.

In addition, people are not only drawn to groups for protection or belonging. Groups can also provide a sense of self-worth and self-esteem. However, this depends on the group being perceived as having value or status—not only by its own members but also, crucially, by others. Here, radical groups face a challenge. They are rarely regarded as high-status groups, at least not in the eyes of the broader society. Individuals who join such groups therefore risk losing status rather than gaining it. Many radical groups appear to be aware of this and have developed various strategies to counteract it.

For example, if a radical group recognises that it is unlikely to gain acceptance within mainstream society, it may instead seek recognition within a narrower community. An Islamist group, for instance, might attempt to enhance its status among Muslims in Europe rather than among the European majority population. Some groups go further by actively distancing themselves from mainstream values and constructing an alternative value system. One example is certain Salafist groups in Britain, which claim to restore a “pure” and authentic form of religious practice. Another strategy involves elevating the moral standing of the in-group by discrediting others. In such cases, the group not only presents itself as superior according to its own criteria but also attributes negative characteristics to out-groups.

To summarise, from the perspective of motivational theory, radical groups can become attractive under certain conditions. Resourceful groups may be appealing if they are able to take over functions normally associated with the state, such as providing security and basic social services. They can also offer community, belonging, and, in some cases, a sense of status. If individuals are drawn to radical groups for these reasons, this attraction can be understood as a result of normal psychosocial processes.

Finally, I would like to offer a brief personal reflection. In many parts of Europe, anti-immigration parties have gained political ground. Immigration, particularly Muslim immigration, has often been framed as a problem, and immigrants themselves as a burden. This rhetoric has increasingly been echoed not only by fringe actors but also by mainstream political parties.

From the perspective of Muslim immigrants, such a social climate may foster a sense of low status in the eyes of the majority society, and in some cases a feeling of rejection. Under these conditions, the appeal of radical groups may increase. Their message of belonging—of being welcomed, valued, and respected within a Muslim community—can be highly attractive. Likewise, claims that “we” possess a superior moral or religious status may resonate. In this sense, the nationalism promoted by anti-immigration movements may, paradoxically, prove counterproductive. The sense of exclusion it generates can be readily exploited by radical groups.

Since the early days of research on violent extremism back in the 1970s and even earlier, one of the main research questions has been why people become extremists or terrorists. However, the flip side of the question was not asked: why and how do violent extremists quit? For a long time, it was taken for granted that these activists remained involved in violent extremism until they ended up dead or in prison. There was hardly any interest in whether some of them possibly disengaged from violent extremism in other ways – even if such knowledge would be of great relevance for prevention and intervention of violent extremism.  

When I worked on my doctoral dissertation on “Racist and Right-Wing Violence in Scandinavia” (Bjørgo, 1997) during the mid-1990s, I interviewed many active, but increasingly also many former activists from the neo-Nazi scenes. I soon realized that the common wisdom that “once a Nazi, always a Nazi” was simply not true. Many participants – including some of those who had been highly ideologically committed and deeply involved in violence – had eventually left the movement and tried to find a foothold in “normal” society. Some succeeded quite well in reintegrating with jobs and families, whereas others struggled and ended up at the margins of society. 

Searching for better understanding of processes of disengagement, deradicalisation and reintegration has been a main research topic throughout my academic career. As a pioneer, I have contributed in establishing this field on the international research agenda. Together with my colleague John Horgan, we developed a much-used conceptual framework with the distinction between deradicalisation and disengagement as two linked – but loosely linked – processes (Horgan & Bjørgo, 2009). I also introduced push, pull and barrier factors to understand processes of change into and out of extremist groups (Bjørgo, 1997; 2011; 2025). Based on my doctoral research, I established the very first Exit project in Norway in 1997, a model that inspired similar Exit projects in Sweden, Germany and eventually in many other countries. During the last 20 years, deradicalisation and disengagement have become main goals in P/CVE policies.

Interviewing former extremists has been a main source of data in my own research in understanding processes of engagement and radicalisation, and deradicalisation, disengagement and reintegration. Many researchers and students want to get access to former extremists to make their own interviews, but few formers are willing, often because they find it stressful to have to recount again and again traumatic and shameful parts of their lives. Together with Robert Örell – himself a former who has led Exit programmes and Exit training for almost 20 years – we decided to use our extensive networks of formers to make a collection of interviews that could be made available to other researchers, students and P/CVE practitioners. We interviewed and developed a dataset with transcribed interviews (at present 15 but growing) for researchers and students, as well as a series of podcast interviews available to the public. 

Read more about the Formers project, the dataset and the podcasts here: 
https://www.sv.uio.no/c-rex/english/resources/formers/index.html

To hear more about my own story on working with former extremists, please listen to my podcast: 
https://www.sv.uio.no/c-rex/english/research/podcast/episodes/podd_toreb.mp3

References:

Bjørgo, T. (1997). Racist and Right-Wing Violence in Scandinavia: Patterns, Perpetrators and responses. Oslo: Tano Aschehoug.

Bjørgo, T. (2011). Dreams and disillusionment: Engagement in and disengagement from militant extremist groups. Crime, Law and Social Change. Volume 55, Issue 4 (2011), pp. 277-285.

Bjørgo, T. (2025). Conceptualising diversity among violent extremists – a typology and a model for explaining change. In: M. Obaidi & J.R. Kunst (eds.). Cambridge Handbook of the Psychology of Violent Extremism, Cambridge University Press.

Bjørgo, T. & Horgan, J. (2009). Leaving Terrorism Behind: Individual and Collective Disengagement. London: Routledge.