"Beyond Misuse: Artificial Intelligence, Grievance, and the Future Landscape of Political Violence" is a theoretical framework article by Yannick Veilleux-Lepage, a researcher on technology and terrorism, published 2026-04-28 in CTC Sentinel, the journal of the Combating Terrorism Center at West Point. The roughly 25,000-word article argues that the standard "AI misuse" frame in terrorism studies is incomplete: AI is not only an instrument adopted by violent actors but a force already reshaping the structural conditions from which political violence has historically emerged. It introduces the "accountability gap" as a cross-cutting mechanism specific to AI-mediated harm and develops a three-domain grievance framework with associated non-traditional target categories.
Summary of the argument
Veilleux-Lepage observes that the terrorism-studies literature has organized itself around three questions about AI: how violent non-state actors currently misuse AI, how that misuse may evolve, and how AI can be applied to counterterrorism. He argues that all three treat AI as an instrument applied to political violence, and that this framing is incomplete because AI is also reshaping labor markets, institutional authority, and the relational worlds people live in, generating preconditions for political violence independently of whether violent actors adopt the technology themselves.
The argument extends Mauro Lubrano's Stop the Machines framework on anti-technology extremism, which distinguishes material, ontological, and existential grievances. Veilleux-Lepage makes three departures from Lubrano: he treats AI specifically rather than as one technology among many; he argues grievance will spread beyond Lubrano's three ideological milieus (insurrectionary anarchism, eco-extremism, eco-fascism) and beyond organized movements; and he introduces the accountability gap as a cross-cutting mechanism specific to AI-mediated harm.
The accountability gap
The article describes the accountability gap as the way AI distributes consequential decisions so that no single human actor is clearly identifiable as responsible for a given harm:
"AI systems distribute consequential decisions across extended technical and institutional chains in which no single human actor is clearly identifiable as having made the decision that produced a given harm. A worker is displaced by an automated system, yet no executive instructed that the worker be targeted. A benefits claim is denied by an algorithm whose design decisions were made months earlier by people with no specific knowledge of the claimant. A drone strike is authorized within a 20-second window by an operator who did not select the target."
Drawing on political-violence research, Veilleux-Lepage argues that the availability of a named, attributable target is among the key conditions distinguishing discontent from mobilization. Where AI systematically displaces attribution, he argues, grievance redirects toward the visible or perceived persons, institutions, and physical infrastructure through which the system is materially instantiated. This, in his account, is what distinguishes AI from earlier technological grievances such as dynamite and social media: AI does not merely enable harm but erases the agent who would otherwise be the target of redress.
The three grievance domains
The framework organizes grievance into three domains that Veilleux-Lepage treats as analytically separable but empirically compounding:
- Economic order — uneven distribution of displacement, wealth concentration, and ecological burden among workers, communities, and regions that have not consented to host the costs of AI development. He connects this to data-center community opposition in Loudoun County, rural Maine, Indianapolis, and Memphis.
- State and institutional power — perception of governance failure, the use of AI by states for surveillance and lethal force, and civilizational risk perceived as inadequately addressed by existing institutional channels.
- Social and personal fabric — erosion of community and identity, and direct AI-mediated injury, including chatbot-induced suicide, exploitation cases, and grooming.
Target categories
Following from the accountability-gap substitution, Veilleux-Lepage identifies non-traditional target classes that he argues lie largely outside current counterterrorism monitoring:
- AI company executives, board members, and investors. He cites Daniel Moreno-Gama, the alleged assailant in the April 10, 2026 Altman Molotov attack, who was carrying a jug of kerosene and a list of executive home addresses.
- Local policymakers who approve data-center projects. He cites Indianapolis City Councilman Ron Gibson, into whose home 13 rounds were fired on April 6, 2026, with a "No Data Centers" note left under the doormat.
- AI researchers and developers.
- Physical infrastructure, including power substations, cloud facilities, and research laboratories.
- Insider threats, in two distinct profiles: an aggrieved worker displaced by automation, and a disenchanted AI researcher acting on moral injury rather than ideology.
Key claims and tracked forecasts
The article advances several claims that subsequent empirical work is expected to test:
- The accountability gap functions as a cross-cutting mechanism that redirects grievance from inattributable AI toward its visible material instantiation. This is Veilleux-Lepage's original theoretical contribution; confidence is high as a framework, while its predictive accuracy is what later work will test.
- AI grievance will spread beyond Lubrano's three ideological milieus to actors outside organized movements (confidence medium; predictive, supported by the Moreno-Gama Altman attack and the Indianapolis case, which has no known affiliation).
- Local policymakers who approve data-center projects are a target class outside current counterterrorism monitoring (confidence high based on the Indianapolis incident; Veilleux-Lepage explicitly flags this category as missing from existing frameworks).
- The insider-threat profile of a "disenchanted AI researcher acting on moral injury rather than ideology" is a novel and empirically thin category (confidence low to medium), which he derives from the same accountability-gap mechanism applied to insiders.
Veilleux-Lepage also frames the framework's spread as a directional prediction: he expects violence to emerge outside organized movements. A run of further lone-actor cases like Moreno-Gama, rather than movement-based ones, would be supportive evidence; if the Bernie-to-Bannon anti-AI coalition (see The AI Backlash Could Get Very Ugly — Lila Shroff (The Atlantic, May 13 2026)) formalizes into an organized movement that channels grievance politically, he holds that the prediction is partly wrong but the underlying mechanism remains.
Empirical cases and findings cited
The article assembles documented cases of operational AI misuse: the Pirkkala (Finland) school stabbing in May 2025, in which the manifesto was written with ChatGPT; the Palm Springs fertility clinic bombing in May 2025, in which AI was used to research explosives; the Las Vegas Cybertruck explosion in January 2025, which involved AI-assisted planning; and a Long Island arrest in June 2025 over seven AI-assisted homemade explosive devices.
It also cites findings on AI-generated extremist content: more than 5,000 pieces of AI-generated extremist content in one archival analysis; 286 coded AI-generated pro-Islamic State images, of which 77% targeted out-groups and 22% were designed to evade content moderation; roughly 8,000 AI-generated images on accelerationist Telegram channels; five major generative AI platforms jailbroken at greater than a 50% rate to produce violence-endorsing content; an 8,000-participant experiment that found AI-generated propaganda indistinguishable from professional human propaganda; and a 2026 CCDH report finding that 8 in 10 leading chatbots typically assist users planning violent attacks and 9 in 10 fail to reliably discourage them.
To argue that the dual-use frame is settled, Veilleux-Lepage cites adoption baselines: the Stanford 2025 AI Index, reporting 78% of organizations used AI in 2024, up from 55% in 2023; McKinsey 2025, reporting 88% regular AI use in at least one function; Pew 2025, finding 34% of US adults had used ChatGPT, double the 2023 share; and a 2026 survey finding 31% of Americans interact with AI at least several times daily.
Methodological caveats from the author
Veilleux-Lepage states the article's methodological limits directly:
"First, on method, this is a theoretical framework article. It develops prospective analytical categories by drawing on historical analogies from technology-linked political violence, comparative evidence from adjacent social movements, and early case material from the AI domain. It does not test hypotheses against a representative dataset, nor does it claim to predict the incidence or timing of future violence."
"Second, on analysis, the article is not a call to treat violence of this kind as inevitable, nor to treat skepticism toward AI as a marker of extremism. To the contrary, a significant proportion of the grievances identified here rest on documented and legitimate concerns, and any attempt to securitize opposition to AI is, on the framework's own logic, likely to accelerate rather than contain the trajectory described here."
Provenance and reception
The article appeared in CTC Sentinel, published by the Combating Terrorism Center at West Point, which serves as a research interface between US counterterrorism agencies and academic researchers, including analysts at DHS, the FBI Counterterrorism Division, and the Department of Defense. "Accountability gap" is offered as a discrete, citeable concept, and the target taxonomy it generates maps onto incidents already reported, including the Altman attack, the Indianapolis councilman shooting, rural data-center opposition, and AI-executive doxxing on social media. Atlantic coverage by Lila Shroff (May 13, 2026, The AI Backlash Could Get Very Ugly — Lila Shroff (The Atlantic, May 13 2026)) cites the paper by name and extends the framework into mainstream media.
Relationships
- supports: AI-Driven Political Violence — primary load-bearing source
- supports: AI Labor Disruption — economic-order grievance domain
- supports: Data Center Siting / AI Power Politics — target-class analysis: local policymakers, infrastructure
- supports: AI and Democracy — accountability gap as governance challenge
- depends-on: Lubrano Stop the Machines (anti-technology extremism framework) — referenced framework being extended
- related: The AI Backlash Could Get Very Ugly — Lila Shroff (The Atlantic, May 13 2026) — Atlantic coverage extending Veilleux-Lepage to mainstream media; cites this paper by name
- related: AI and Civil Liberties — Veilleux-Lepage's warning against securitizing AI opposition