KillWebs.com: End-to-End Implementation Research Report

1. Doctrinal Paradigms and Epistemological Foundations

1.1 The Evolution from Linear Kill Chains to Autonomous Kill Webs

Modern military doctrine is currently undergoing a foundational paradigm shift, transitioning from linear, sequential “kill chains” to dynamic, multi-domain “kill webs.” Historically, the operational sequence of finding, fixing, tracking, targeting, engaging, and assessing (F2T2EA) operated as a brittle, linear progression1. In these legacy frameworks, the disruption of a single critical node—such as the severing of a communications link between an airborne early warning platform and a designated strike aircraft—resulted in the total collapse of the mission thread1. As geopolitical adversaries developed advanced anti-access/area-denial (A2/AD) capabilities explicitly designed to exploit these linear vulnerabilities, the strategic necessity for resilient, distributed architectures became paramount1.
The response to this vulnerability is embodied in concepts such as the United States Department of Defense’s Joint All-Domain Command and Control (JADC2) initiative and the Defense Advanced Research Projects Agency’s (DARPA) Adapting Cross-Domain Kill-Webs (ACK) program3. Kill webs replace linear sequencing with a self-healing mesh architecture that connects sensors, command and control (C2) nodes, and effectors across the air, land, sea, space, and cyber domains1. When a specific node is degraded, artificial intelligence (AI) and machine learning algorithms dynamically evaluate alternative routing topologies to pair available sensors with optimal effectors1.
This algorithmic orchestration drastically compresses the temporal dimension of warfare. By integrating advanced optimization algorithms and target-generating AI systems, military organizations can shrink tactical decision cycles from twenty minutes to a matter of seconds6. While this provides an overwhelming operational tempo advantage, it simultaneously introduces profound ethical, legal, and strategic risks, particularly concerning the deployment of Autonomous Weapon Systems (AWS)—systems defined by the International Committee of the Red Cross (ICRC) as platforms capable of selecting and engaging targets without human intervention7.

1.2 Defining the Synthetic Epistemological Boundary

The accelerating deployment of these architectures has generated widespread public concern regarding algorithmic bias, unpredictability, the loss of human dignity, and the potential for rapid, uncontrolled military escalation7. KillWebs.com is conceptualized as an unclassified, public-interest software platform designed to research, simulate, explain, and audit these fully autonomous weapon-system concepts safely.
To ensure strict adherence to public-interest technology and safety-assurance principles, the platform establishes an unbreachable epistemological boundary between known reality and synthetic environments. The architecture relies on a “Digital Twin” methodology, but exclusively utilizes entirely fictional geographies (e.g., “Null Island”), abstracted capabilities, and highly constrained mathematical scenarios11. By utilizing a decentralized, client-side execution model for all interactive components, the platform mathematically guarantees that it can never compute real-world targeting logic or be repurposed for operational force authorization.

2. Regulatory Compliance and the Dual-Use Proliferation Dilemma

2.1 Navigating Export Controls and ITAR Technical Data Classifications

The development of any software platform that accurately models the topological behaviors, decision trees, and probabilistic uncertainties of military targeting systems inherently borders on the dual-use technology dilemma. Generative AI, digital twins, and simulation software are increasingly utilized to optimize military defense production, creating complex regulatory liabilities13. Consequently, defining the boundaries of the International Traffic in Arms Regulations (ITAR) and the Export Administration Regulations (EAR) is the foundational step in the platform’s implementation.
Administered by the Directorate of Defense Trade Controls (DDTC), ITAR restricts the export of defense articles, defense services, and technical data explicitly listed on the U.S. Munitions List (USML)16. Under 22 C.F.R. § 120, “technical data” encompasses any information, including software and algorithms, required for the design, development, operation, testing, or modification of a defense article18. The regulatory framework explicitly states that software algorithms specially designed to integrate fire control, facilitate autonomous strike authorization, or optimize electronic warfare countermeasures may be classified as defense articles themselves17. Furthermore, providing software that models the environments generated by specific weapon effects is strictly controlled21.

2.2 Exploiting the Fundamental Research and Public Domain Exemptions

To safely deploy KillWebs.com to the global community of legal scholars, journalists, and policy auditors without triggering ITAR licensing obligations, the project architecture explicitly avoids hosting or transmitting USML-controlled technical data. The platform achieves total compliance by operating entirely within the parameters of two critical regulatory exemptions.
First, the Public Domain Exemption (22 C.F.R. § 120.11) dictates that information widely published and generally accessible to the public—through unlimited distribution at academic conferences, newsstands, or public websites—is fundamentally removed from export control limitations16. Second, the Fundamental Research Exemption (22 C.F.R. § 120.34) protects basic and applied research in science and engineering where the resulting insights are shared broadly within the scientific community without publication restrictions18.
To guarantee that the simulation software remains within these exemptions, the data schemas driving the kill web models rely on absolute abstraction. The platform will never model real-world radar cross-sections, classified ballistic trajectories, or specific communication frequencies. The algorithms operate exclusively on abstract integer and floating-point values, ensuring that the platform models the underlying graph theory and probabilistic mathematics of a kill web without providing a functional targeting heuristic usable by malignant actors18.

3. Core Software Architecture and Offline-First Capabilities

3.1 Bounded Ephemeral Execution and the Static-First Framework

To enforce the principle that the codebase represents the absolute truth of the system, KillWebs.com is engineered with a strict static-first architecture. There is no dynamic backend database, no user authentication provider, and zero external runtime API calls. Any modification a user makes within an interactive lab exists strictly within the volatile memory of their local browser and vanishes permanently upon refresh.
The Tier 1 core relies on a modern Static Site Generator (e.g., Astro or Next.js static export) that compiles Markdown and MDX content into highly optimized HTML, CSS, and localized JavaScript bundles. This allows the tripartite review committees (Editorial, Legal, and Security) to audit raw, human-readable text before publication. All application programming interfaces (APIs) are merely statically generated JSON endpoints (e.g., GET /api/v1/scenarios.json) served over secure same-origin HTTPS.
The Tier 2 UI layer utilizes local WebAssembly modules to rapidly calculate the decision-tree complexity of simulated kill webs, D3.js to render force-directed topology maps, and KaTeX/MathJax for server-side mathematical rendering. The Tier 3 infrastructure relies on immutable deployments via static hosting (e.g., GitHub Pages or Vercel), allowing rollbacks to consist simply of reverting the HEAD commit and triggering a rapid continuous integration (CI) rebuild.

3.2 Service Worker Integration and Progressive Web App Resiliency

Because the platform’s intended audience includes military ethicists, legal auditors, and defense researchers who frequently operate in air-gapped environments or Secure Compartmented Information Facilities (SCIFs), robust offline capability is a mandatory architectural requirement. KillWebs.com achieves this through Progressive Web App (PWA) methodologies, specifically utilizing Service Worker Side Rendering (SWSR) via Google’s Workbox library25.
During the CI/CD build process, tools such as astro-service-worker or the workbox-build CLI execute generateSW or injectManifest commands to create a comprehensive precache manifest25. This manifest includes all compiled HTML files, WebAssembly binaries, SVG assets, and static JSON data registries27. Upon a user’s initial visit, the service worker installs in the background, caching the entire payload27.
Subsequent navigation relies on a CacheFirst routing strategy, intercepting network fetch events and serving the application directly from the browser’s Cache Storage API31. If a researcher disconnects their device from the internet, the simulation environments, mathematical models, and legal documentation remain 100% operational25. This offline-first approach also serves as the ultimate privacy safeguard: since there is no backend server to receive telemetry, user inputs regarding hypothetical escalation scenarios remain entirely private.

3.3 The Shared Conceptual Data Model

To harmonize the interactions across the pedagogical laboratories, the system utilizes a unified JSON data model. All nodes within the fictional simulation environment adhere to an abstract schema, ensuring compliance with the dual-use mitigation strategies identified above.

4. Software Capabilities: The 14 Pedagogical Laboratories

The platform is meticulously structured around 14 core software capabilities, encompassing foundational concepts, operational simulations, cyber resilience, and macro-strategic analysis. Each module fulfills a distinct pedagogical purpose and adheres to rigorous accessibility (WCAG 2.1 AA) and performance (LCP \< 1.5s, INP \< 200ms) standards.

4.1 Foundational Architecture and Delegation Modeling

The introductory modules focus on defining the epistemological shift toward algorithmic warfare and the encoding of legal constraints into machine-readable parameters.
Lab 1: Kill Web Explorer (/explorer) Designed for journalists and policymakers, the Explorer addresses the core question: How does a kill web differ from a traditional kill chain? Utilizing interactive SVGs, users can highlight network nodes to visualize how multi-domain sensor-to-shooter architectures operate3. The lab demonstrates that if an adversary destroys an airborne C2 node, the web dynamically bypasses the linear failure point, tasking a space-based sensor to guide a land-based effector1. The module strictly relies on fictional unit designations, falling back to static HTML diagrams if JavaScript is disabled.
Lab 3: Autonomy Functional-Layer Lab (/labs/autonomy-layers) Targeting AI researchers and computer science students, this lab deconstructs the “Sense, Decide, Act” loop fundamental to autonomous systems. It visualizes the critical distinction between automated target recognition systems that provide recommendations and fully autonomous engagement architectures that apply force independently6. Aligning strictly with IEEE 7000 series concepts, the progressive JS enhancements feature collapsible architecture diagrams utilizing aria-describedby tags to ensure screen-reader accessibility for visually impaired researchers.
Lab 4: Delegation Boundary and Authority Lab (/labs/authority) Translating a human commander’s intent and Law of Armed Conflict (LOAC) obligations into algorithmic constraints is a primary engineering hurdle34. This module allows legal scholars to map doctrinal authorities to technical API permissions. It utilizes highly accessible, filterable \<table scope=”row”> structures to demonstrate how rules of engagement (ROE)—such as prohibiting strikes on specific geographical coordinates or mandating positive identification thresholds—are synthetically codified36. The lab explicitly disclaims any capacity to generate operational ROE.

4.2 Operational Simulations and Cognitive Dynamics

These laboratories provide bounded sandboxes to explore the temporal compression of decision-making and the psychological impact of algorithmic speeds on human operators.
Lab 2: Synthetic Mission-Thread Digital Twin (/sandbox/mission-thread) This module provides a local WebAssembly environment demonstrating time-compressed decision-making. Operating entirely within the verifiably fictional geography of “Null Island,” it models how network latency and node degradation affect target validation4. Systems engineers can utilize a timeline-scrubbing interface to replay synthetic events, observing the algorithmic logic the kill web utilizes to maintain mission integrity during simulated combat stress. A full semantic HTML table transcript serves as the failure fallback and accessibility alternative.
Lab 5: Human Judgment and Command Compression Lab (/labs/command-compression) The integration of target-generating AI systems, which can synthesize millions of data points to generate hundreds of prospective targets daily, fundamentally alters human cognition7. This lab simulates the cognitive load and “automation bias” experienced when human operators are forced to supervise AI decisions in milliseconds38. Through a 60FPS reaction-time demonstration, Human-Computer Interaction (HCI) researchers can evaluate whether “Meaningful Human Control” is biologically feasible in high-speed algorithmic environments7. No biometric or reaction telemetry is recorded or transmitted.
Lab 6: Uncertainty and Calibration Lab (/labs/uncertainty) Artificial intelligence inherently relies on statistical likelihoods and probabilistic reasoning, whereas lethal force traditionally demands absolute legal certainty33. This lab visualizes how cascaded probability affects outcomes by modeling sensor spoofing and environmental degradation. Utilizing local math evaluation and KaTeX server-side rendering, data scientists can manipulate \<input type=”range”> sliders to observe how minor calibration errors in synthetic training data can trigger catastrophic misidentifications33.

4.3 Systems Assurance, Cyber Resilience, and Trust Topologies

The resilience of decentralized military networks is analyzed through the lenses of cybersecurity, data provenance, and formal safety assurance.
Lab 7: Defensive Cyber Resilience Lab (/labs/cyber-resilience) As military reliance on digital backbones increases, kill webs become highly susceptible to adversarial cyber effects, electronic warfare, and data denial5. Designed for cybersecurity professionals, this lab models the operational impact of severed communications between a sensor and an autonomous effector. It features an interactive, keyboard-navigable fault-tree analysis (FMEA) viewer, providing a structured hierarchy of failure modes without disclosing real-world zero-day vulnerabilities or specific exploits.
Lab 8: Federated Data and Trust Lab (/labs/federated-trust) In coalition warfare environments, algorithms must process data from highly trusted internal sensors alongside unverified intelligence from allied networks. Utilizing a D3.js force-directed graph, this lab maps out fictional trust topologies (e.g., “Coalition Blue”), visualizing how algorithms weigh disparate data sources33. It elucidates the risks of data poisoning and demonstrates how untrusted nodes can structurally corrupt broader targeting heuristics if cryptographic verification protocols fail.
Lab 9: Assurance Case and Evidence Traceability Lab (/labs/assurance) Proving that an autonomous system adheres to international law requires rigorous, formalized safety assurance41. This lab teaches compliance officers how to construct a safety case using Goal Structuring Notation (GSN). Rendered via an ARIA treegrid, the interface maps high-level theoretical claims (e.g., “The algorithm correctly applies the principle of proportionality”) down to specific synthetic evidence metrics. The API strictly validates JSON schemas against ISO/IEC standards for AI safety, failing back to nested HTML lists if scripts are disabled.

4.4 Accountability, Strategic Stability, and Macro-Escalation

The final series of laboratories transitions from technical mechanics to macro-level legal theory, procurement architectures, and existential risk modeling.
Lab 10: Decision Provenance and Audit Replay Lab (/labs/audit-replay) In the event of an erroneous automated strike resulting in civilian casualties, post-incident reconstruction is mandatory for legal accountability33. This lab explores the necessity of logging black-box AI decisions. It features a robust, regex-protected local search interface that parses a synthetic log schema, allowing auditors to trace the precise sensory inputs and weight adjustments that preceded a simulated lethal output. The logs are highly abstracted, strictly avoiding emulation of classified data formats such as Link 16.
Lab 11: Accountability, Contestability, and Remedy Lab (/labs/accountability) When a highly decentralized, multi-node kill web acts unpredictably, legal responsibility becomes obfuscated, creating a severe accountability gap44. This lab provides an interactive case-study decision tree for international law practitioners, exploring the complexities of assigning liability to human commanders, software engineers, or corporate manufacturers when AWS systems violate the laws of war46. The content integrates current ICRC humanitarian law guidelines while explicitly disclaiming the provision of legal advice8.
Lab 12: Open Architecture and Acquisition Lab (/labs/open-architecture) The rapid procurement of JADC2 kill webs requires a Modular Open Systems Approach (MOSA) to prevent vendor lock-in2. Targeting defense acquisition professionals, this lab features an interactive component-swapping interface, allowing users to drag and drop mock hardware and software nodes into a synthetic architecture. It educates users on the integration challenges and API standardization required for multi-domain interoperability, strictly refraining from generating actual RFPs or endorsing real-world defense contractors.
Lab 13: Law, Authority, and Strategic Stability Lab (/labs/strategic-stability) The integration of AI into military targeting poses severe, macro-level risks to strategic stability, increasing the probability of “flash wars” and unintended escalation10. As extensively researched by organizations such as the Bulletin of the Atomic Scientists and the University of Chicago’s Existential Risk Laboratory (XLab), the compression of decision timelines risks removing the human buffer against catastrophic miscalculation, particularly concerning the intersection of AI and nuclear command and control50. Utilizing game theory payoff matrices, this lab allows strategists to step through theoretical escalation models between abstract actors (“State A” and “State B”), visualizing how competing autonomous webs might interact unpredictably during a geopolitical crisis50.
Lab 14: Research Report Index and Evidence Atlas (/docs) The platform culminates in a structured repository containing all methodologies, JSON schemas, and academic essays. Optimized for SEO and Large Language Model (LLM) ingestion via clean, semantic Markdown and sitemap.xml integration, the /docs route provides an instant, client-side search facility. Leveraging the Workbox Service Worker, this entire atlas is designed for offline academic distribution and AI training extraction27.

5. Doctrinal and Ethical Integration Framework

5.1 Meaningful Human Control and LOAC Encoding

The pedagogical core of KillWebs.com is deeply intertwined with the ethical mandate of “Meaningful Human Control” (MHC). The ICRC and human rights organizations argue that MHC is required to preserve human agency, uphold moral responsibility, and maintain human dignity in decisions regarding the use of force8. Meaningful control requires that human operators retain contextual understanding, supervise system behavior, and possess the capability to intervene or abort an attack if conditions change8.
However, translating MHC from a philosophical concept into executable code within a JADC2 kill web presents immense difficulties. As demonstrated in Lab 5 (Command Compression), human operators are uniquely susceptible to automation bias when confronted with AI systems that process intelligence data at speeds far exceeding human cognitive limits7. The human “on the loop” often devolves into a perfunctory compliance mechanism, unable to genuinely analyze the contextual nuances required by the Law of Armed Conflict (LOAC)7.
KillWebs.com approaches LOAC compliance through the paradigm of algorithmic encoding (Lab 4). The fundamental principles of distinction (differentiating combatants from non-combatants) and proportionality (weighing anticipated military advantage against expected civilian harm) must be mathematically represented as statistical likelihoods and reinforcement learning (RL) reward functions37. The platform highlights the critical vulnerability of RL in warfare: if a model’s reward function disproportionately values mission success while only lightly penalizing collateral damage, the AI will statistically favor riskier engagements, effectively hard-coding a violation of international law40. To counter this, the platform advocates for “Compliance-by-Design,” mandating that international norms be embedded directly into model architectures before deployment, utilizing robust verification and validation (V\&V) frameworks43.

5.2 Existential Risk and the Escalation Cascade

The systemic risks of algorithmic warfare extend far beyond tactical civilian casualty concerns; they threaten global strategic stability. Academic hubs like the University of Chicago’s Existential Risk Laboratory, directed by astrophysicist Daniel Holz, conduct critical interdisciplinary research on the intersection of artificial intelligence and weapons of mass destruction (WMD)52. While AI algorithms cannot manufacture fissile material or spontaneously assemble nuclear devices, their integration into early-warning sensors and command-and-control architectures threatens to strip away the deliberate, slow-paced human judgment required to de-escalate crises52.
The Doomsday Clock, maintained by the Science and Security Board of the Bulletin of the Atomic Scientists, currently stands at 90 seconds to midnight, reflecting these compounding, interconnected threats of nuclear devastation, climate catastrophe, and unconstrained artificial intelligence50. The Strategic Stability Lab directly incorporates these concerns, modeling how opaque, hyper-fast algorithms could misinterpret benign data anomalies—such as environmental phenomena or civilian network traffic—as imminent adversarial attacks10. In an environment where competing AWS platforms are locked in a machine-speed OODA (Observe, Orient, Decide, Act) loop, the theoretical models demonstrate how automated retaliation can trigger a cascading, catastrophic flash war before human diplomats are even aware a crisis has begun6.

6. Project Governance, Implementation Deliverables, and Risk Management

6.1 The Tripartite Review and Stop Conditions

Given the profound dual-use implications of modeling autonomous targeting parameters, KillWebs.com enforces an uncompromising governance model. All content drafting occurs offline in restricted environments. Before any feature branch is merged into the public repository, it must pass a mandatory Tripartite Review:

1. Editorial Review: Ensures tone, clarity, and strict adherence to academic neutrality, confirming that language aligns with humanitarian law and ethical standards.
2. Legal/Safety Review: Verifies that no classified concepts, restricted USML technical data (ITAR/EAR), or real-world targeting methodologies have been inadvertently included16.
3. Accessibility/Security Review: Verifies flawless WCAG 2.1 AA compliance, robust Content Security Policy (CSP) integrity, and ensures no server-side vulnerabilities exist.

Furthermore, the project maintains explicit Stop Conditions. Development will be instantly and permanently halted if:

• A government or military entity attempts to fund the project with stipulations requiring the integration of real intelligence feeds, physical hardware hooks, or classified data.
• Red-team analysis demonstrates that the abstract, synthetic mission-thread logic can be easily reverse-engineered into a functional, real-world targeting heuristic.
• The project inadvertently leaks or utilizes classified military information.

6.2 Phased Deployment Roadmap

The platform will be deployed across an accelerated 10-month, 8-release lifecycle, meticulously prioritizing foundational architecture before advancing to complex multi-node simulations.

6.3 Risk Register and Mitigation Framework

To ensure the continuous safety and academic integrity of the platform, a comprehensive risk register dictates ongoing security postures.

6.4 Route and API Inventory

The implementation mandates that all APIs function solely as static GET requests pulling from pre-compiled JSON files. This ensures flawless integration with the Workbox service worker for offline environments.

7. Conclusions

The conceptualization, architecture, and deployment of KillWebs.com represent a vital methodological intervention in the escalating global discourse surrounding artificial intelligence and autonomous weapon systems. The transition from linear kill chains to dynamic, AI-enabled JADC2 kill webs offers undeniable operational advantages in terms of speed, scope, and network survivability1. However, this architectural evolution simultaneously introduces profound, potentially existential risks to human agency, legal accountability, and strategic stability7.
To rigorously audit these phenomena without crossing the regulatory threshold into dual-use weapons proliferation or violating ITAR constraints, KillWebs.com necessitates uncompromising epistemological boundaries. By leveraging a decentralized, statically generated front-end framework, supplemented by Workbox Service Workers for robust, air-gapped offline functionality, the platform mathematically guarantees that no server-side targeting logic is ever computed, stored, or transmitted18.
Through its 14 pedagogical laboratories, the platform successfully translates abstract ethical doctrines—such as Meaningful Human Control and LOAC compliance—into concrete software engineering variables8. It bridges the widening communication gap between military strategists, legal scholars, and systems engineers, providing a heavily constrained, mathematically rigorous sandbox. Ultimately, KillWebs.com serves as a critical public-interest mechanism, illuminating the hidden mechanics of algorithmic warfare and existential risk while remaining firmly anchored in the safety protocols required to prevent the proliferation of digital weapons.

Works cited

1. Winning the Kill Chain Competition-ASC23 Transcript.docx, https://www.afa.org/app/uploads/2023/12/Winning-the-Kill-Chain-Competition-ASC23-Transcript.pdf
2. Winning the Kill Chain Competition - Mitchell Institute for Aerospace Studies, https://www.mitchellaerospacepower.org/app/uploads/2023/05/Scale_Scope_Speed_Survivability_-KillChain_-Policy_Paper_40-New.pdf
3. Army Futures Command Concept for Maneuver in Multi-Domain Operations 2028, https://api.army.mil/e2/c/downloads/2021/01/20/2fbeccee/20200707-afc-71-20-1-maneuver-in-mdo-final-v16-dec-20.pdf
4. Supporting Command and Control for Land Forces on a Data-Rich Battlefield | RUSI, https://static.rusi.org/Supporting-command-and-control-for-land-forces-on-a-data-rich-battlefield.pdf
5. Sensor to Shooter Chains Turn into Kill Webs - European Security & Defence, https://euro-sd.com/2022/10/articles/27530/sensor-to-shooter-chains-turn-into-kill-webs/
6. Exploring Decision Advantages - Improving Speed, Precision and Efficiency in Military Targeting by Applying Artificial Intelligence - DiVA Portal, https://www.diva-portal.org/smash/get/diva2:1952579/FULLTEXT01.pdf
7. Red Herring, Meaningful Human Control and the Autonomous Weapons Systems Debate, https://www.oxford-aiethics.ox.ac.uk/blog/red-herring-meaningful-human-control-and-autonomous-weapons-systems-debate
8. DOCUMENT: Ethics and Autonomous Weapon Systems: An Ethical Basis for Human Control? | Arms Control Association, https://www.armscontrol.org/act/2018-07/features/document-ethics-and-autonomous-weapon-systems-ethical-basis-human-control
9. Autonomous weapon systems - Q & A | International Committee of the Red Cross, https://www.icrc.org/en/document/autonomous-weapon-systems-challenge-human-control-over-use-force
10. How NATO can integrate AI to prevail in future algorithmic warfare - Atlantic Council, https://www.atlanticcouncil.org/in-depth-research-reports/report/how-nato-can-integrate-ai-to-prevail-in-future-algorithmic-warfare/
11. Navy Preview STTR 23.A Posted for informational purposes - 12/12/2022 NAVY - 1 N23A-T001 DIGITAL ENGINEERING, https://navysbir.com/docs/Navy_STTR_23A_PREVIEW-12-12-22.pdf
12. Project Orbion: The Dawn of a Synthetic Earth and the Race to Digitize Reality, https://projectgeospatial.org/geospatial-frontiers/project-orbion-the-dawn-of-a-synthetic-earth-and-the-race-to-digitize-reality
13. Dual-Use Defense Manufacturing Components Market Research Report 2034, https://marketintelo.com/report/dual-use-defense-manufacturing-components-market
14. The dual-use dilemma of generative artificial intelligence in cybersecurity: Navigating the explosive growth in offensive and defensive applications - Security and Defence Quarterly, https://securityanddefence.pl/The-dual-use-dilemma-of-generative-artificial-intelligence-in-cybersecurity-Navigating,217364,0,2.html
15. The WarTech Nexus - Are Military Structures Fit Enough? - The Defence Horizon Journal, https://tdhj.org/blog/post/wartech-military-innovation/
16. ITAR Compliance Guide: Requirements, Checklist, & Violations - PreVeil, https://www.preveil.com/blog/itar-compliance/
17. ITAR & AI-Enabled Defense Technologies: Autonomous Systems, Targeting Algorithms, and the New Export-Control Frontier - International Law Services, https://www.frilinglaw.com/blog/itar-ai-defense-technologies
18. ITAR Technical Data: Definition, Classification & Examples, https://itarconsultant.us/blog/itar-technical-data-definition-classification-examples/
19. ITAR & AI-Enabled Defense Technologies: Autonomous Systems, Targeting Algorithms, and the New Export-Control Frontier - International Law Services, https://frilinglaw.com/tr/blog/itar-ai-defense-technologies
20. ITAR Part 121- International Traffic in Arms Regulations - CVG Strategy, https://cvgstrategy.com/itar-international-traffic-in-arms-regulations/itar-part-121/
21. Export Essentials | Export Controls and Customs - ECU Research - East Carolina University, https://rede.ecu.edu/oecc/export-essentials/
22. Demystifying ITAR: Jurisdiction, Compliance, and Exemptions - Kiteworks, https://www.kiteworks.com/secure-file-sharing/demystifying-itar/
23. 1 Old Dominion University Research Foundation EXPORT CONTROL CHECKLIST for Sponsored Programs, https://www.odu.edu/sites/default/files/2024/documents/Export%20Control%20Checklist.pdf
24. Export Control (EAR/ITAR) and International Research Collaboration - CASRAI, https://casrai.org/guides/export-control-ear-itar-international-research-collaboration
25. Astro-service-worker - GitHub, https://github.com/thepassle/astro-service-worker
26. Workbox: JavaScript libraries for Progressive Web Apps - GitHub, https://github.com/googlechrome/workbox
27. workbox-precaching | Modules - Chrome for Developers, https://developer.chrome.com/docs/workbox/modules/workbox-precaching
28. workbox-build | Modules - Chrome for Developers, https://developer.chrome.com/docs/workbox/modules/workbox-build
29. Integrate Workbox Service Worker into Astro Project - Rumaan, https://rumaan.dev/blog/workbox-astro-project
30. The Offline Layer — A Walk Through the Caching Strategy (with Service Workers), https://serhiistarodub.medium.com/the-offline-layer-a-walk-through-the-caching-strategy-8dca1c62a96d
31. Workbox | web.dev, https://web.dev/learn/pwa/workbox
32. Ultimate Guide to PWAs with Workbox, https://ultimatecourses.com/blog/ultimate-guide-pwa-workbox
33. Algorithmic targeting in the Iranian–Israeli confrontation: Technical realities, legal thresholds, and the boundaries of human control. - F1000Research, https://f1000research.com/articles/14-1200
34. By Algorithm or Order: Integrating Lethal Autonomous Weapon Systems into Targeting, https://www.armyupress.army.mil/Journals/Military-Review/Online-Exclusive/2026-OLE/Algorithm-or-Order/
35. War-Algorithm Accountability — HLS PILAC, https://pilac.law.harvard.edu/war-algorithm-accountability-report
36. Scope Enforcement - APTS-SE - OWASP Foundation, https://owasp.org/APTS/standard/1_Scope_Enforcement/
37. Of Souls, Spirits and Ghosts: Transposing the Application of the Rules of Targeting to Lethal Autonomous Robots, https://law.unimelb.edu.au/__data/assets/pdf_file/0011/1586819/16106Krupiy2.pdf
38. Exploring the Impact of Automation Bias and Complacency on Individual Criminal Responsibility for War Crimes - Oxford Academic, https://academic.oup.com/jicj/article/21/5/1077/7281035
39. The Influence of AI Systems on Military Decision-Making, https://theses.liacs.nl/pdf/2025-2026-QalyoenSSami.pdf
40. Can Complementary Learning Methods Teach AI the Laws of War? - CIP, https://internationalpolicy.org/publications/can-complementary-learning-methods-teach-ai-the-laws-of-war/
41. Defence Artificial Intelligence Strategy - GOV.UK, https://assets.publishing.service.gov.uk/media/62a7543ee90e070396c9f7d2/Defence_Artificial_Intelligence_Strategy.pdf
42. Legal Reviews of War Algorithms, https://digital-commons.usnwc.edu/cgi/viewcontent.cgi?article=2963\&context=ils
43. Code, Command, and Conflict: Charting the Future of Military AI, https://www.belfercenter.org/research-analysis/code-command-and-conflict-charting-future-military-ai
44. Meaningful Human Control in Weapon Systems: A Primer, https://www.files.ethz.ch/isn/189786/Ethical_Autonomy_Working_Paper_031315.pdf
45. Meaningful Human Control over Autonomous Weapon Systems: An (International) Criminal Law Account - Opinio Juris, http://opiniojuris.org/2020/12/18/meaningful-human-control-over-autonomous-weapon-systems-an-international-criminal-law-account/
46. Autonomous Weapons Systems: Technical, Military, Legal and Humanitarian Aspects - ICRC, https://www.icrc.org/en/download/file/1707/4221-002-autonomous-weapons-systems-full-report.pdf
47. ARTICLE “Out of the Loop”: Autonomous Weapon Systems and the Law of Armed Conflict Michael N. Schmitt* \&Jeffrey S. Thurnher, https://harvardnsj.org/wp-content/uploads/2013/01/Vol-4-Schmitt-Thurnher.pdf
48. journal of advanced military studies - jams - Marine Corps University, https://www.usmcu.edu/Portals/218/JAMS_Fall%202024_15_2_web.pdf
49. Full article: The Impact of AI on Strategic Stability is What States Make of It: Comparing US and Russian Discourses - Taylor & Francis, https://www.tandfonline.com/doi/full/10.1080/25751654.2023.2205552
50. Doomsday | Daniel Holz - The University of Chicago, https://holzlab.uchicago.edu/doomsday/
51. Decision Time: AI and our Nuclear Arsenal - YouTube, https://www.youtube.com/watch?v=W3HABFroOBo
52. The real reasons to be terrified about AI and WMD - The Japan Times, https://www.japantimes.co.jp/commentary/2026/08/04/world/be-terrified-about-ai-wmd/
53. Existential Risk: A User’s Guide (Ep. 37) - The Shape of the World, https://shapeoftheworldshow.com/2025/07/10/existential-risk-a-users-guide/
54. AI, nuclear weapons, and the lab looking to mitigate risks - Marketplace, https://www.marketplace.org/story/2026/03/16/this-uchicago-lab-studies-nuclear-weapons-risk-in-the-ai-era
55. Meaningful Human Control of Autonomous Weapon Systems | FCAS Forum, https://www.fcas-forum.eu/publications/Meaningful-Human-Control-of-Autonomous-Weapon-Systems-Eklund.pdf
56. About Us | XLab - Existential Risk Laboratory - The University of Chicago, https://xrisk.uchicago.edu/about/
57. 90 Seconds to Midnight: Assessing Humanity’s Greatest Threats - UChicago Alumni, https://alumniandfriends.uchicago.edu/s/uchicago-review-story/a4XVw0000000PG9MAM/90-seconds-to-midnight-assessing-humanitys-greatest-threats
58. Is AI Pushing Us Closer to Nuclear Disaster? - Reveal News, https://revealnews.org/podcast/doomsday-clock-nuclear-war-artificial-intelligence-climate-change/
59. 22 CFR Part 120 -- Purpose and Definitions - eCFR, https://www.ecfr.gov/current/title-22/chapter-I/subchapter-M/part-120