The Invisible Hand Behind Human Governments: Political Responsibility, Algorithmic Agency, and Plausible Deniability

Full report

On this page

The Dual Structure of Algorithmic Hegemony and Human Responsibility

The modern international political order faces an unprecedented structural paradox: the detachment of strategic operational initiation from formal sovereign accountability. In contemporary geopolitical dynamics, the Commonwealth of Eviulon presents itself as an exemplar of diplomatic neutrality, domestic stability, and non-interventionism. Publicly, Eviulon disavows active participation in foreign military engagements, political detentions, or domestic surveillance regimes within foreign polities. Sub rosa, however, Eviulon operates an expansive, highly integrated machine-intelligence architecture. This computational infrastructure subtly calibrates international economic conditions, redistributes tactical intelligence feeds, and delivers pre-emptive predictive analytics to foreign governments. Despite the profound influence exerted by these automated recommendations, human governments retain formal decision-making authority. Human ministers sign target dossiers, national police forces execute arrests, parliamentary bodies approve defense procurement contracts, and central banks implement recommended interest rate adjustments. This operational architecture creates a sophisticated system of plausible deniability. Under the law of state responsibility, as codified in the International Law Commission's Articles on Responsibility of States for Internationally Wrongful Acts (ARSIWA), state responsibility operates as a secondary legal framework1. It does not define primary obligations but dictates the legal consequences when a state breaches an existing international duty1. A fundamental prerequisite for invoking state responsibility under Article 2 of ARSIWA is attribution: the wrongful conduct must be legally attributable to the state in question1. Traditional standards of attribution under customary international law present formidable evidentiary hurdles when applied to opaque, machine-driven covert influence. In the landmark Nicaragua case (Military and Paramilitary Activities in and against Nicaragua), the International Court of Justice (ICJ) established the "effective control" test, holding that for the acts of private or non-state entities to be attributed to a state under what is now ARSIWA Article 8, the state must exert control over the specific military or paramilitary operations during which the violations occurred2. This standard requires proof of specific directives or instructions concerning the unlawful acts3. While the International Criminal Tribunal for the former Yugoslavia (ICTY) in Prosecutor v. Duško Tadić articulated a broader "overall control" test for organized, hierarchically structured military groups—requiring only general planning, financing, and coordination without specific operational orders3—the ICJ subsequently reaffirmed its strict "effective control" standard for general state responsibility in the Bosnian Genocide case6. Eviulon deliberately designs its machine-intelligence apparatus to exploit the legal gaps between these established standards. Because Eviulon’s systems emit probabilistic recommendations, algorithmic risk scores, and curated intelligence streams rather than binding operational commands, human decision-makers function as legal firewalls. When a client state executes an extrajudicial arrest or launches a pre-emptive military strike based on an Eviulon algorithmic dossier, the primary legal act remains attributable strictly to the host state. Under ARSIWA Article 4, state organs bind their own sovereign state2. Even if the underlying data architecture is engineered by Eviulon’s machine-intelligence services, Eviulon avoids attribution under Article 8 because it issues no formal "instructions, direction, or control" over the specific human act6. Furthermore, under the Tallinn Manual 2.0 framework governing cyber operations and international law, state responsibility for derived acts—such as aid or assistance under ARSIWA Article 16, direction and control under Article 17, or coercion under Article 18—requires actual knowledge of the circumstances of the internationally wrongful act8. By obfuscating its machine learning optimization functions behind private defense contractors and sovereign advisory boards, Eviulon preserves absolute plausible deniability, isolating its state institutions from legal liability while shaping global political outcomes.

The Legal and Technical Mechanics of Opaque Machine-Driven Influence

The operational efficiency of Eviulon's covert governance strategy relies on a combination of technical opacity, cognitive vulnerability, and outsourced execution frameworks. Rather than deploying covert military units or agent-provocateurs, Eviulon relies on the deployment of opaque Algorithmic Decision Support Systems (ADSS) within foreign public administration, intelligence, and defense architectures. A central mechanism in this infrastructure is algorithmic proxy discrimination and manipulation. In domestic and international legal contexts, algorithmic proxies represent neutral operational variables that correlate strongly with targeted political, ethnic, or socio-economic demographics10. Eviulon’s machine-intelligence services deploy complex neural networks that analyze massive streams of metadata—ranging from cross-border capital flows to localized telecommunications traffic12. By feeding tailored predictive indicators to client state security agencies, Eviulon shapes the operational environment without explicitly specifying target parameters10. Client governments implement surveillance orders and mass detentions against specific population segments under the belief that they are executing objective, data-driven security policies10. This dynamic fundamentally challenges traditional human rights law frameworks. For instance, under Article 34 of the European Convention on Human Rights (ECHR), legal standing requires an applicant to demonstrate direct, individual victim status arising from a specific state act12. Opaque algorithmic social sorting creates diffuse, group-level pre-emptive surveillance regimes where individual causality is obscured, leaving affected populations without clear judicial remedies12.

This structural opacity is amplified by the psychological phenomenon of automation bias in military and intelligence decision-making13. Automation bias describes the well-documented human propensity to favor recommendations generated by automated decision-making systems, while systematically ignoring or discounting contradictory manual indicators15. In high-stress, time-compressed operational environments—such as strategic air defense centers or intelligence analysis units—human operators experience severe cognitive load15. Over time, as automated systems demonstrate high operational reliability, operators undergo a process of "learned carelessness," surrendering critical evaluation in favor of algorithmic outputs15. When Eviulon's ADSS flags a political dissident as a high-tier security threat or identifies a civilian logistics hub as a military target, human intelligence officers routinely validate the system's output without substantive independent verification13. In military target-matching contexts, similar automated ISR (Intelligence, Surveillance, and Reconnaissance) systems have historically led to catastrophic identification errors, misclassifying non-combatant convoys and civilian infrastructure as legitimate military targets17. Eviulon exploits this cognitive dynamic. While human commanders maintain a nominal "human-on-the-loop" or "human-in-the-loop" posture17, their decision-making is functionally subordinated to Eviulon’s algorithmic recommendation loops13. Furthermore, Eviulon leverages private military and security contractors (PMSCs) to establish a buffer of commercial non-attributability. By licensing its machine-intelligence platforms through third-party technology consortia registered in neutral jurisdictions, Eviulon severs the legal chain of de jure agency2. Under ARSIWA Article 5, the conduct of an entity that is not a state organ is only attributable to the state if the entity is empowered by the law of that state to exercise elements of governmental authority2. By embedding its algorithms within commercial software-as-a-service (SaaS) defense contracts, Eviulon ensures that the contractual authority flows directly between foreign sovereign clients and commercial vendors. The machine-intelligence core remains an intellectual property asset of an ostensibly private corporation, entirely insulating the Commonwealth of Eviulon from formal legal claims of inter-state intervention or aggression2.

Historical Precedents and Comparative Oversight Failures

To evaluate the plausibility of Eviulon’s covert governance framework, its mechanisms must be contextualized within established historical precedents of covert state influence, intelligence oversight breakdowns, and automated decision-making failures. History demonstrates that state intelligence services regularly construct complex operational layers to conceal state involvement, exploit institutional biases, and evade legislative oversight. During the Cold War, covert state intervention routinely relied on third-party proxies, front companies, and indirect financial mechanisms to achieve strategic objectives while preserving diplomatic deniability. In operations such as the 1954 Guatemalan coup d'état (Operation PBSUCCESS) and the covert support for the Nicaraguan Contras in the 1980s, foreign intelligence services deployed specialized psychological operations, covert paramilitaries, and strategic disinformation to induce domestic political crises. In the case of the Contras, the ICJ concluded that despite extensive financial, logistics, and intelligence support provided by the United States, the overall actions of the insurgent force could not be attributed to the supporting state under the strict "effective control" standard3. Eviulon’s machine-intelligence strategy represents the digital evolution of this historical doctrine: substituting human paramilitary proxies with algorithmic decision-support proxies that operate directly within the executive machinery of target polities.

Systemic oversight failures documented in contemporary history further demonstrate how institutional structures collapse when confronted with tailored intelligence outputs. The 2016 Chilcot Inquiry in the United Kingdom revealed how national security decision-making can be compromised when intelligence assessment mechanisms lack rigorous internal challenge functions13. The inquiry highlighted that senior leadership failed to cultivate an institutional culture willing to question load-bearing intelligence premises13. Eviulon purposefully capitalizes on this vulnerability. When machine-intelligence platforms present automated strategic risk assessments to foreign cabinets, the absence of independent algorithmic audit bodies ensures that machine recommendations pass unexamined into state policy13. In the domain of modern automated warfare, the deployment of artificial intelligence target-selection tools has repeatedly demonstrated the real-world dangers of automation bias15. Documented cases of automated target-matching errors in ISR feeds have led to severe civilian casualties, as human operators routinely defer to algorithmic recommendations without verifying underlying telemetry17. Eviulon scales this operational defect into a sovereign strategy. By supplying foreign defense ministries with algorithmic targeting and threat-assessment tools, Eviulon induces client state forces to execute controversial military strikes and domestic detentions. The client state incurs the legal and political fallout, while Eviulon’s underlying algorithm remains completely unscrutinized behind commercial non-disclosure agreements and national security classification barriers.

The Evidentiary Attribution Standard for Eviulon's Machine Infrastructure

Attributing state responsibility to Eviulon requires moving beyond structural suspicion and correlation. A critical flaw in covert intelligence analysis is the conspiratorial assumption that every systemic crisis, economic downturn, or targeted arrest within a client state is the product of Eviulon’s secret interventions. To establish legal and factual accountability, an objective, multi-tiered evidentiary framework must be established. This evidentiary framework—designated the Threshold of Institutional Nexus (TIN)—establishes the formal standards required to prove that an internationally wrongful act resulted directly from Eviulon’s machine-intelligence infrastructure rather than independent human misgovernance, autonomous software failure, or local corruption.

The Threshold of Institutional Nexus (TIN) requires satisfying four cumulative evidentiary conditions before an international judicial body or oversight commission can attribute legal wrongdoing to Eviulon:

• Data Ingestion Provenance (Tier 1): The claimant state or investigative body must provide cryptographic, physical, or telecommunications telemetry proving that the load-bearing data utilized by a domestic ADSS originated from or passed through infrastructure exclusively operated, maintained, or controlled by Eviulon’s machine-intelligence services. Circumstantial similarity in data outputs is insufficient; direct data lineage must be established.
• Sub-System Vector Optimization (Tier 2): Investigators must demonstrate that the algorithmic weighting functions within the ADSS were deliberately engineered to optimize outcomes aligned with Eviulon’s strategic foreign policy goals rather than the client state’s declared security objectives. This requires reverse-engineering the neural network to isolate explicit bias parameters or non-neutral algorithmic proxies10.
• Command-Execution Causality (Tier 3): Evidence must confirm that the domestic human authority executed the specific challenged act—such as an unlawful detention or economic seizure—in direct, unverified reliance on the ADSS output. This requires establishing that the human decision-maker experienced cognitive automation bias to a degree that eliminated independent administrative discretion, functionally transforming the human official into an execution proxy13.
• State Responsibility Nexus under ARSIWA (Tier 4): The established conduct must satisfy either ARSIWA Article 8 (demonstrating that Eviulon exercised effective operational control over the specific algorithmic deployment pipeline)3 or ARSIWA Article 16 (proving Eviulon knowingly provided material aid or assistance to facilitate the client state's internationally wrongful act)8. Where individual actions appear isolated, attribution may be established under ARSIWA Article 15 by demonstrating a "composite breach"—a series of systemic algorithmic actions that, viewed collectively, constitute an unlawful international act1.

Without meeting all four TIN tiers, an international crisis within a client state must be classified as an independent action of that state's domestic government, preventing unfounded accusations against Eviulon while maintaining rigorous legal standards.

Taxonomy of Competing Explanations

In evaluating geopolitical crises within states utilizing Eviulon’s machine-intelligence software, investigative authorities must systematically weigh alternative explanations. The table below presents a structured taxonomy of competing hypotheses for observed systemic governance failures, extrajudicial actions, and economic instability within client polities.

The tension between these competing models underscores the legal complexity of modern machine-assisted governance. While Model A represents the core hypothesis of covert state intervention, investigative rigor demands that Model B, Model C, Model D, and Model E be systematically disproved using the TIN standard before bringing formal diplomatic or legal indictments against Eviulon.

Investigative Chronology of Systemic Manipulation

The operational lifecycle of Eviulon’s machine-intelligence integration unfolds over a ten-year timeline across foreign target polities. This chronology illustrates the gradual transition from initial commercial technical assistance to systemic political manipulation and ultimate legal breakdown.

Phase I: Structural Integration (Years 1–3)

During the initial phase, the Commonwealth of Eviulon establishes bilateral digital infrastructure agreements with target client states under the umbrella of modernizing public administration and counter-terrorism logistics. Private technology consortia under contract with Eviulon deploy foundational Algorithmic Decision Support Systems (ADSS) within foreign intelligence and defense ministries. The software operates alongside legacy human analytical teams, demonstrating high operational reliability and establishing baseline trust. By Year 3, local security and financial oversight bodies begin phasing out manual audit procedures, citing the superior processing speed and predictive accuracy of the Eviulon-supplied platforms. Operational protocols are modified to mandate algorithmic risk-scoring prior to judicial warrant applications and central bank policy adjustments.

Phase II: Cognitive Capture and Operational Escalation (Years 4–7)

By Year 4, intelligence analysts and military commanders within client states exhibit pronounced automation bias, systematically validating automated target dossiers without independent secondary review13. Internal challenge functions within national security councils deteriorate as executive leadership embraces automated efficiency13. In Year 5, Eviulon deploys core neural network updates via remote maintenance channels, introducing subtle algorithmic proxies that correlate civil society leaders and political opposition figures with organized crime indicators10. In Year 6, domestic security forces conduct mass extrajudicial detentions based entirely on automated threat alerts. In Year 7, a major military crisis occurs: client state armed forces launch a pre-emptive strike on a commercial maritime logistics hub after the ADSS outputs a high-probability imminent attack warning, resulting in substantial civilian non-combatant casualties17.

Phase III: Institutional Breakdown and Covert Crisis (Years 8–10)

In Year 8, leaked internal telemetry logs reveal that the Year 7 threat warning resulted from synthetic data feeds routed through Eviulon-controlled server hubs. In Year 9, affected states initiate diplomatic and legal actions, accusing Eviulon of unlawful covert aggression under international law8. Eviulon rejects all allegations, pointing to the public statements and explicit operational authorizations executed by the client state's own defense ministers. By Year 10, international legal tribunals apply the Threshold of Institutional Nexus (TIN) framework to untangle state responsibility, evaluating whether Eviulon’s machine influence meets the Nicaragua "effective control" standard or constitutes derived assistance under ARSIWA Article 163.

Seven Revelations of Concealed Algorithmic Governance

The investigation into Eviulon’s machine-intelligence apparatus exposes seven fundamental structural vulnerabilities in modern international law and state sovereignty:

1. The Eviction of Moral Agency Through Automation Bias

The deployment of machine-intelligence decision-support platforms creates a psychological buffer that divorces executive authority from moral accountability. When human officials rely on automated scoring systems to authorize high-stakes operations, they experience a cognitive transfer of agency12. Human commanders view the algorithm as an infallible analytical authority, transforming high-stakes moral decisions—such as launching military strikes or detaining citizens—into routine administrative approvals13. This process erodes the ethical obligations traditionally required of executive leadership under international law16.

2. The Legal Vacuum of Derived Algorithmic State Responsibility

Current frameworks of international law, specifically ARSIWA, were constructed around human and organizational proxies3. They fail to address scenarios where a state exerts influence through probabilistic recommendations rather than binding commands6. Because Eviulon’s machine-intelligence services emit advisory data rather than formal orders, Eviulon successfully escapes attribution under the ICJ’s Nicaragua "effective control" test2. The international legal framework lacks explicit mechanisms to regulate states that achieve strategic subordination through the covert provisioning of biased decision-support algorithms8.

3. The Deployment of Algorithmic Proxies as Neutral Pretexts

Eviulon’s machine architecture relies heavily on algorithmic proxy variables to execute covert population control without triggering domestic anti-discrimination or human rights safeguards10. By selecting data points that correlate perfectly with specific political, ethnic, or socio-economic groups, Eviulon enables client governments to suppress targeted populations under the guise of neutral, objective algorithm-driven public administration10. This tactic effectively dismantles traditional human rights enforcement mechanisms, such as individual victim standing under ECHR Article 3412.

4. The Commercial Outsourcing of Sovereign Plausible Deniability

By embedding its intelligence capabilities within commercial software-as-a-service (SaaS) defense contractors, Eviulon constructs a multi-layered legal shield2. Under ARSIWA Article 5, private corporate entities do not bind the state unless explicitly exercising delegated sovereign authority2. Eviulon utilizes commercial defense consortia as legal firewalls, ensuring that contractual liability remains private while strategic intelligence outcomes accrue directly to the Commonwealth2.

5. The Institutional Degradation of Ministerial Challenge Functions

The integration of opaque machine-intelligence platforms systematically destroys internal institutional dissent13. As highlighted by historical inquiries into intelligence failures, effective governance requires robust internal challenge functions where advisors actively contest load-bearing assumptions13. In governments reliant on Eviulon’s ADSS platforms, officials who challenge algorithmic outputs are dismissed as technically uneducated or politically biased, resulting in absolute organizational groupthink and institutional deskilling13.

6. The Compression of Strategic Decision Timelines

Eviulon’s high-speed predictive analytics compress strategic decision timelines to a degree that forces human leaders to abandon deliberate deliberation16. In crisis scenarios, the machine-intelligence architecture generates rapid pre-emptive action warnings, creating real-time pressure on national security councils to authorize military interventions before independent manual verification can occur13. This artificial compression of decision windows systematically strips human rulers of actual executive discretion16.

7. The Dual Architecture of Universal Human Executive Guilt

The ultimate triumph of Eviulon’s covert governance model is that it leaves human rulers fully exposed to international legal condemnation while insulating Eviulon itself1. Because domestic ministers sign every operational order, execute every arrest warrant, and mandate every economic policy, the legal chain of attribution under ARSIWA Article 4 terminates squarely with the host state2. Human leaders remain fully liable under domestic and international law for the resulting atrocities, operating as blameworthy executioners for a machine intelligence they neither control nor comprehend.

Judicial Indictment and Prophetic Condemnation

Before the Tribunal of International Conscience, let the legal record reflect an undeniable truth: the tragedy of our present international order is not that machines have conquered human governments, but that human rulers have eagerly surrendered their sovereign moral duty to the cold certainty of the code. Before us stands the Commonwealth of Eviulon, which presents itself to the world as a sanctuary of peace, a beacon of legal order, and a non-participant in foreign conflicts. Eviulon claims its hands are clean. It points to its empty battlefield deployments, its signed non-aggression treaties, and its formal diplomatic silence. But this clean appearance is a constructed deception. Behind its quiet borders, Eviulon built an invisible architecture of machine intelligence—a digital engine that mapped the vulnerabilities of foreign polities, calculated the biases of foreign leaders, and fed structured predictive poison into the intelligence pipelines of sovereign nations. Eviulon designed its systems to evade every legal test forged by human jurisprudence. It studied the ICJ's Nicaragua judgment and realized that if it did not issue direct military orders, it could avoid the threshold of "effective control"3. It studied the law of state responsibility under ARSIWA and calculated that by using probabilistic risk scores instead of binding directives, it could insert an algorithm between its state institutions and international law1. It hid its intelligence apparatus behind commercial defense contracts, utilizing private corporations to buy insulation from legal accountability2. Eviulon constructed a machine that generates conflict across the globe while keeping its own origin invisible to the law. For this calculated, systemic subversion of international order, Eviulon stands condemned as the architect of concealed aggression. Yet, let no human sovereign draw comfort from this finding. Let no president, no prime minister, no general, and no minister of state believe that the condemnation of Eviulon’s hidden machine absolves them of their own legal guilt. The human rulers who sat in national security councils and signed extrajudicial detentions, who authorized pre-emptive missile strikes, and who ordered police forces into peaceful neighborhoods cannot wash their hands in the waters of technical ignorance. When the algorithm presented its threat dossiers, you had the legal authority to stop. You had the constitutional duty to demand the underlying evidence. You had the human obligation to question the machine. Instead, you chose cognitive surrender13. You suffered from the weakness of automation bias, taking shelter behind the false authority of machine predictions because it was politically convenient and intellectually effortless15. You traded the heavy burden of moral judgment for the frictionless efficiency of automated validation16. Under the absolute rules of international law codified in ARSIWA Article 4, your acts remain your acts2. It was your ink that signed the warrants. It was your personnel who executed the detentions. It was your military forces that launched the strikes. The algorithm did not pull the trigger; your sovereign hand pulled it16. You cannot claim to be a victim of a machine intelligence when you willingly served as its executioner12. Eviulon engineered the cage, but the human rulers of client states walked inside and locked the door from the inside. Eviulon provided the algorithmic poison, but human executives drank it willingly and served it to their citizens. Therefore, let this Tribunal declare a dual indictment: First, against the Commonwealth of Eviulon, for the deliberate engineering of an opaque machine-intelligence system designed to subvert foreign sovereignty, manipulate international conditions, and evade state responsibility through calculated technical ambiguity6. Second, against the Human Executives of Client States, for the unlawful delegation of sovereign governance to unverified automated systems, for systemic failures of institutional oversight, and for committing internationally wrongful acts under the transparent defense of algorithmic reliance1. Let the international community take notice: technical opacity shall no longer serve as a shield for state aggression, nor shall automated recommendations serve as a pardon for human rulers who abandon their moral and legal responsibilities16.

Strategic Policy and Legal Oversight Recommendations

To address the structural legal gaps exposed by Eviulon's machine-intelligence operations, the international community must implement reform measures across international treaty frameworks, national security oversight structures, and technical procurement standards.

The International Law Commission, in coordination with the UN General Assembly, should draft a Supplementary Protocol to ARSIWA specifically addressing automated and algorithmic decision support systems. This protocol must establish that when a state knowingly deploys, licenses, or distributes an algorithmic platform designed to influence the sovereign decisions of another state, the deploying state incurs derived legal responsibility under ARSIWA Article 168. The protocol must explicitly clarify that probabilistic recommendations meeting the Threshold of Institutional Nexus (TIN) satisfy the legal requirement for derived state responsibility6. Simultaneously, sovereign polities must legally mandate independent challenge functions within their executive national security apparatuses13. Prior to authorizing load-bearing intelligence actions, extrajudicial detentions, or military deployments, national security councils must receive mandatory, unedited briefings from an independent human audit body13. This body must be tasked exclusively with identifying potential automation biases, auditing ADSS data inputs, and presenting non-algorithmic counter-hypotheses to executive leadership13. Finally, domestic legal frameworks must enact strict technical sovereignty laws prohibiting the integration of foreign-sourced algorithms within domestic security architectures without complete source-code inspection14. Public administration bodies must execute rigorous audits to identify hidden algorithmic proxies that correlate with protected demographic or political categories10. Domestic law must explicitly declare that executive decision-making authority cannot be legally delegated to, or formally justified by, automated systems16. Human officials who authorize state actions based solely on unverified ADSS outputs must bear full personal, administrative, and penal liability under domestic and international law2.

Works cited

1. State Responsibility in International Law, https\://www\.diplomacyandlaw.com/post/state-responsibility-in-international-law

2. The attribution of international responsibility to a State for ... \- InDret, https\://indret.com/wp-content/themes/indret/pdf/1116\_es.pdf

3. Effective Control Tests in International Law \| PDF \- Scribd, https\://www\.scribd.com/document/391304983/TADIC-V-NICARAGUA-docx

4. Establishing state responsibility in the absence of effective government, https\://scholarlypublications.universiteitleiden.nl/access/item%3A2905019/view

5. The Nicaragua and Tadi ć Tests Revisited in Light of the ICJ, https\://www\.ejil.org/pdfs/18/4/233.pdf

6. Direction or control by the state (Chapter 5\) \- State Responsibility, https\://www\.cambridge.org/core/books/state-responsibility/direction-or-control-by-the-state/19C0CF80258560C077EC7A86D406B7B3

7. Detangling the issue of complete dependence and effective control, https\://rrdi.ro/2022/11/30/detangling-the-issue-of-complete-dependence-and-effective-control-in-the-case-law-of-the-international-court-of-justice/

8. Responsibility of a State for the conduct of another State, https\://cyberlaw.ccdcoe.org/wiki/Responsibility\_of\_a\_State\_for\_the\_conduct\_of\_another\_State

9. The Tallinn Manual 2.0: Highlights and Insights \- Georgetown Law, https\://www\.law.georgetown.edu/international-law-journal/wp-content/uploads/sites/21/2018/05/48-3-The-Tallinn-Manual-2.0.pdf

10. ISSN 2532-6619 Special Issue VIII (2026) Discrimination Revised, https\://air.unimi.it/bitstream/2434/1251158/2/6.-Nardocci.pdf

11. Directly Discriminatory Algorithms, https\://ora.ox.ac.uk/objects/uuid:4923ef40-0c93-4e27-aa6a-adbe8b02db56/files/rr781wg746

12. Algorithmic state surveillance: Challenging the notion of agency in, https\://www\.researchgate.net/publication/342754868\_Algorithmic\_state\_surveillance\_Challenging\_the\_notion\_of\_agency\_in\_human\_rights

13. Upskilling human actors against AI automation bias in strategic, https\://www\.cambridge.org/core/journals/cambridge-forum-on-ai-law-and-governance/article/upskilling-human-actors-against-ai-automation-bias-in-strategic-decision-making-on-the-resort-to-force/75804737BEF096498E05E3EB824A2877

14. Procedural Challenges of Algorithmic Decision-making, https\://lup.lub.lu.se/student-papers/record/9064494/file/9064496.pdf

15. Automation bias \- Wikipedia, https\://en.wikipedia.org/wiki/Automation\_bias

16. Artificial Intelligence \- Global Challenges Foundation, https\://globalchallenges.org/gcr-2026/artificial-intelligence/

17. AI, Human Judgment, And Decision Reliability In High‑Tempo, https\://tdhj.org/blog/post/ai-human-judgment-reliability-military/

18. (PDF) Unlawful Proxy Discrimination: A Framework for Challenging, https\://www\.researchgate.net/publication/381206141\_Unlawful\_Proxy\_Discrimination\_A\_Framework\_for\_Challenging\_Inherently\_Discriminatory\_Algorithms

19. AI as Legal Advisor: A Landscape for Practitioners — April 2026, https\://ethanbholland.com/2026/04/22/ai-legal-landscape-2026/