Faster Than the Loop: Closing the ADF’s AI Human Control Gap

Shani Edwards, Department of Defence, GLOSEC 2026

Executive Summary

Australia’s strategic environment is increasingly characterised by persistent cyber competition, with the emergence of artificial intelligence (AI) as a critical component of future warfare. The 2026 National Defence Strategy and the Australian Signals Directorate 2024-2025 Threat Report indicate that hostile actors are increasingly using AI-enabled capabilities leveraging machine-speed (Australian Signals Directorate, 2025). In order to ensure the Australian Defence Force (ADF) is prepared to fight in an AI-enabled conflict, it must ensure Military Commanders are able to operationalise AI-enabled systems to effectively neutralise or destroy hostile threats (Department of Defence, 2026). While the Department of Defence’s (Defence’s) Policy Settings for Responsible Use of Artificial Intelligence in Defence provide a strong framework for the ethical, lawful and accountable use of AI, it falls short in providing realistic and distinguishable guidance for AI control model application (Department of Defence, 2026). It fails to recognise that reversible non-kinetic outcomes (such as AI-enabled cyber network defence) and irreversible kinetic outcomes (such as an AI-enabled strike package) require different layers of risk acceptance to be delivered to Military Commanders. As the ADF continues to explore AI-enabled capabilities, such as autonomous cyber defence systems and counter-uncrewed aerial systems (C-UAS), clearer policy guidance is required to discriminate against human control requirements. 

This policy paper concludes that the most immediate governance gap that exists is related to Control Model application for AI-enabled systems. The existing Defence policy applies the same generic Control Models used for cyber defence and administrative applications as it does for AI-enabled weapons, despite the substantially higher legal and operational risks. The policy does not adequately account for circumstances where AI defensive capabilities can respond faster than human intervention, and where this response is reversible (likely a non-kinetic effect) or irreversible (kinetic effect). Consequently, the policy has not aligned AI governance with the demands of future AI-enabled conflict. 

Accordingly, it is recommended that Defence mandate a differentiated kinetic AI Control Model for AI-enabled kinetic effects with irreversible outcomes, requiring a ‘responsible human chain of command and control’ (U.S Department of State, 2023). This reform would protect Australia’s commitments to human chain of command, and strengthen alignment with Article 36 to the Geneva Conventions’ legal review requirements (International Committee of the Red Cross, 2006). Moreover, it recommends reversible AI-enabled systems be placed under appropriate Control Models aligned to their lower legal and operational consequences. This will encourage higher AI adoption and diffusion across the ADF, preparing the force for the realities of modern conflict. 

Problem Identification

The Policy Settings for Responsible Use of Artificial Intelligence in Defence provides a robust foundation for the use of AI within Defence (Department of Defence, 2026). However, it does not adequately address the operational realities of AI-enabled warfare. Specifically, the policy provides limited guidance on governance arrangements for AI versus AI engagements, where autonomous defensive systems interact directly with autonomous offensive systems in contested cyber environments (Kissell, 2026). As these engagements are likely to occur at machine-speed, existing policy with generic model application regarding human oversight and accountability may be insufficient to support timely and effective operational action. Some commentators have contended that the current Defence policy ‘isn’t fit for purpose for unarmed autonomous systems, including those providing non-lethal electromagnetic attack’ due to current accountability mechanisms that ‘will rationally discourage officials from fielding systems whose behaviour they can’t fully predict’ (Kissell, 2026).

Consequently, the current policy settings do not adequately account for circumstances in which autonomous defensive capabilities may be required to respond faster than human operators can intervene, creating a gap between current AI governance and the demands of future cyber conflict. They also do not effectively address where an Accountable Officer can rationalise functions of AI-enabled capabilities where they do not have complete control over a non-kinetic and reversible outcome (Kissell, 2026) (Department of Defence, 2026). Left unaddressed, this gap risks leaving Australia strategically disadvantaged against peer adversaries who can act at greater speed with fewer legal and ethical constraints. The impact will be heightened risk to critical national infrastructure and inability for the ADF to effectively respond to threats in a time of national crisis (Five Eyes Security Agencies, 2026). 

This concern is best highlighted in the governance of AI-enabled kinetic effects, such as AI-enabled Counter-Uncrewed Aerial System (C-UAS) capabilities used to find and kinetically disrupt or destroy hostile threats. The policy’s only specific provision for AI in weapon systems is a requirement to conduct an Article 36 legal review prior to use, consistent with Australia’s obligations under the Geneva Conventions (International Committee of the Red Cross, 2020) (Department of Defence, 2026). Beyond this legal compliance checkpoint, the policy applies the same generic Control Models used for administrative and cyber defence use cases to kinetic engagement decisions. However, it does not distinguish between reversible outcomes, such as a blocked cyber intrusion, and irreversible outcomes, such as a kinetic strike. The absence of a differentiated Control Model for kinetic AI-enabled effects - one that mandates where human chain of command is required - creates uncertainty for Military Commanders and Accountable Officers seeking to deploy AI-enabled systems (Turillazzi, 2026). The responsibility of Accountable Officer is further complexed by traceability and prediction of AI model behaviour, which industry experts have detailed as ‘very difficult if not impossible’ (Santoni de sio et al, 2021). 

Put simply, the policy does not have a clear position on where the limits of AI decision making start and end for Military Commanders (Turillazzi, 2026). The likely result from this is inconsistent risk acceptance across Defence, slower adoption of critical AI capabilities, and greater exposure to unintended harm and legal consequences without clear standards for where and when human control is imposed.

Context

Australia's Cyber Front

Australia’s preparedness for cyber war is undermined by a growing strategic misalignment between the 2023–2030 Australian Cyber Security Strategy (2023-2030 ACSS) (Department of Home Affairs, 2023) and Australia’s current geo-political context described within the 2026 National Defence Strategy (2026 NDS) (Department of Defence, 2026). The 2023-2030 ACSS is the capstone policy for guiding Defence in the cyber domain. The 2026 NDS frames Australia as operating within a ‘dangerous and unpredictable’ environment characterised by intensified deterioration between great power relations and shrinking warning times for regional conflict. Notably, it identifies that ‘international norms against the use of force and coercion are weakening, with more states already engaged in conflict at the start of 2024 than at any point since 1946’ (Department of Defence, 2026). More specifically within the cyber domain, the Australian Signals Directorate (ASD) identified in their 2024-2025 Threat Report that 'state-sponsored cyber actors continue to pose a serious and growing threat to our nation. They target networks operated by Australian governments, critical infrastructure (CI) and businesses for state goals' (Australian Signals Directorate, 2025).

Conversely, the 2023-2030 ACSS remains structured around gradual, long term leadership and resilience building extending towards 2030 (Department of Home Affairs, 2026). The 2023-2030 ACSS Strategy reflects a careful and indirect approach to addressing the risk of hostile state cyber activities. The ACSS and NDS are not entirely contradictory in their goals in ensuring security of Australia’s cyber domain, but they explicitly reflect different levels of urgency about how close Australia may already be to a major cyber enabled geopolitical crisis (Priyandita, 2026). This year, the Five Eyes cyber security agencies released a joint statement directly calling to action the need for urgency in addressing Frontier AI model use for offensive and defensive capabilities (Five Eyes cyber security agencies, 2026). It stresses an ‘immediate’ need to integrate Defensive AI capabilities. This divergence creates a critical preparedness gap for rapidly evolving threats in the cyber domain, particularly the governance of AI models operating at machine-speed and the use of AI systems for kinetic effects (Priyandita, 2026).

Defence's current approach to AI

The emergence of AI-enabled capabilities presents a strategic predicament that is not fully addressed within Defence's current AI governance framework. Advanced AI systems are increasingly capable of cyber operations at great scale, the emergence of which is ‘particularly worrisome’ for global defence establishments as they reduce the time available for human decision-making and challenging traditional approaches to military use of cyber terrain (Whyte, 2020).

In 2025, Defence released Policy Settings for Responsible Use of Artificial Intelligence in Defence;

Responsible use of AI at all stages of the technology lifecycle. This policy requires that all AI-enabled decisions be tied to human accountability (Department of Defence, 2026) and mandates Proportional Control. In practice, this applies Control Models to AI technology. Control Models ‘describe how a user and AI technology interact with each other, how human control is enacted throughout a system’s operations, and how an AI technology interacts with other AI technologies, systems and machines’ (Department of Defence 2026). The policy does not dictate what Control Models must be applied to what types of AI capabilities, nor does it specify how accountability is maintained when a defensive system must act faster than a human operator can meaningfully intervene. The unintended outcome of this could be too much, or too little, control applied to AI technologies, undermining operational effectiveness of the ADF. 

This gap is more acute in the kinetic domain. The proliferation of low-cost uncrewed aerial systems and drone swarms has driven Defence investment in Counter-UAS capabilities that are able to find and kinetically engage multiple aerial threats faster than a human operator (Conroy, 2026), AI-enabled systems will be integral to engage these threats faster than a human operator (Conroy, 2026). Unlike a defensive cyber AI model, a kinetic C-UAS engagement produces an irreversible physical effect and directly engages Australia’s obligations under international humanitarian law, including the principles of distinction and proportionality (Protocol I, 1977). The current policy directs personnel to seek legal advice on whether, and at what stage, an Article 36 legal review under Additional Protocol I to the Geneva Conventions is required for AI in weapon systems (Department of Defence, 2026). Article 36 is a legal review of the lawfulness of new weapons (International Committee of the Red Cross, 2006), but it does not itself govern human control over AI systems making irreversible engagement decisions (Boulanin et al, 2020).  

Defence’s policy of generic Control Models does not fill that gap either. This is a compliance checkpoint rather than an operational Control Model, and the policy’s existing Control Models make no distinction between an AI system defending a network and an AI system authorised to kinetically engage a target (Department of Defence, 2026). This creates a lower level of assurance for kinetic applications than for cyber applications, despite kinetic effects carrying substantially greater consequences. Furthermore, the policy states it implements Australia’s commitments under the 2023 Political Declaration on the Responsible Use of Military AI and Autonomy which is substantially built around the principle of ‘responsible human chain of command and control’ (US Department of State, 2023). However, the policy's Proportionate Control and Control Models section contains no equivalent requirement (Department of Defence, 2026).

Reforming Defence Policy for Machine-Speed Cyber Defence and AI-Enabled Kinetic Effects

As Defence becomes increasingly reliant on AI-enabled capabilities, policy settings must evolve to address the operational realities of AI-enabled conflict in both the cyber and kinetic domains; both effective governance and operational use must form parts of responsible AI use (Santoni de sio et al, 2021). This is particularly important for AI-enabled kinetic effects such as Counter-UAS, where human control must be commensurate with irreversibility of the outcome (Ekelhof, 2019). It is imperative this is balanced with operational requirements to defend Australia and its interests effectively and ethically (Ekelhof, 2019). 

Furthermore, Australia must align its preparedness in the cyber domain to the strategic tempo of the current geopolitical landscape given the acknowledgement from ASD and Defence that cyber operations will form a central component of future conflict (Australian Signals Directorate, 2025). To address the limitations of the Policy Settings for Responsible Use of Artificial Intelligence in Defence in addressing these challenges and meeting preparedness requirements, this paper proposes reform to establish clear accountability arrangements and control model application for AI-enabled systems that meet Military Commanders operational requirements for using these capabilities in Defence of Australia. 

Policy Options

Option 1: Establish a differentiated Kinetic AI Control Model for AI-enabled Kinetic Effects

This proposal would amend the Policy Settings for Responsible Use of Artificial Intelligence in Defence Control Models section to add a Kinetic AI Control Model, distinct from the existing Control Models identified in current policy for administrative and cyber applications, being:

-          Human-In-The-Loop/Human-On-The-Loop

-          Human-Machine Teaming

-          Machine-Machine-Teaming.

This proposal would require any AI-enabled system authorised to kinetically engage a target, such as an AI-enabled C-UAS system engaging a hostile drone or drone swarm, to be required to operate under a model which mandates a positive human authorisation step prior to the kinetic effect and links directly to the existing Article 36 legal review obligation. This distinguishes kinetic engagement decisions (which are irreversible and directly engage international humanitarian law obligations of distinction and proportionality) from reversible cyber and administrative AI applications. In practice, this would give Military Commanders and Accountable Officers a clear, mandated control model with ‘human chain of command and control’ to apply whenever an AI system is authorised to produce a kinetic effect, rather than relying on case by case judgement from the Accountable Officer.

Practically, this would ensure that the correct and proportionate control model is applied to maintain operational effectiveness of the ADF. This would mean lower levels of human control and oversight for administrative applications and responding to offensive AI threats; whereas higher levels of control applied to AI systems with kinetic outcomes. This amendment would also require habitual review as AI technologies continue to develop. 

Administration of the amendment would be overseen by the Defence Artificial Intelligence Centre (DIAC), in partnership with Defence Legal for Article 36 alignment and relevant Military Command for capability employment (eg. Joint Capabilities Group for C-UAS).

Success would be measured against whether this reform establishes a mandated, differentiated control model for AI-enabled kinetic effects, ensuring a positive human authorisation step precedes any AI-enabled kinetic engagement. Additionally, Australia’s commitment to meaningful human control under the 2023 Political Declaration should be reflected in Defence policy rather than left to case by case interpretation. Finally, lower risk AI-enabled capabilities should be employed effectively at scale to defend Australia and its interests, with lower levels of human control as appropriate. Essential to achieving Option 1 is operational rehearsals, to review and ensure effectiveness of Control Models and positive human chain of command authorisation where required to meet the peer threats. This option directly closes the gap identified in Problem Identification between the policy’s generic Control Models and the distinct legal and ethical stakes of kinetic AI-enabled effects, and will give Military Commanders a clear, consistent standard to apply when using AI capabilities.  

Option 2: Mandate red-teaming and machine-speed test certification for AI-enabled cyber defence systems

This proposal would amend the Policy Settings for Responsible Use of Artificial Intelligence in Defence Proportionate Controls to require red-teaming and machine-speed performance certification before an Accountable Officer approves an AI-enabled cyber defence system for deployment. Red-teaming would involve testing each system against simulated offensive AI attacks to identify vulnerabilities and confirm AI model behaviour under contested conditions. Unlike Options 1 and 3, this option would not create new control models or governance arrangements for autonomous machine-speed response. It would mandate that red-teaming is required as part of the ‘Test and Evaluation and Review’ control measure. Practically, this policy change will result in de-risking AI-enabled systems behaving unpredictably in AI-versus-AI engagements, and allowing for faster reaction time to AI-enabled threats with pre-tested human oversight. This will strengthen assurance of Human-In-The-Loop/Human-On-The-Loop, on the basis improved testing is a lower-risk way to improve preparedness, rather than adding further governance requirements to Control Models already in place.

Administration of the amendment would be overseen by DIAC, in partnership with the ASD for red-team expertise. Success would be measured by risk reduction red-teaming activity outcomes, and data collected on the risk of AI-enabled defensive systems behaving unpredictably in AI-versus-AI engagements. This option is the lowest-cost and lowest-risk of the three, and requires no change to existing Control Models. 

Option 3: Establish a Pre-Authorisation Framework for AI-enabled Cyber Defence

This option proposes to add a Pre-Authorisation Framework to Machine-Speed Control Model to the existing Policy Settings for Responsible Use of Artificial Intelligence in Defence, sitting alongside Human-in-the-Loop, Human-on-the-Loop and Machine-Machine Teaming. Under this amendment, the Accountable Officer would define and approve, in advance, the bounded parameters within which a defensive AI system may act autonomously against an AI-enabled threat, for kinetic and non-kinetic outcomes. Accountability for actions taken within these parameters would remain with the Accountable Officer, consistent with current policy position that accountability cannot be transferred to a technology.

Administration and implementation would be overseen by DIAC, including the publishing of pre-authorisation frameworks and conduct of rehearsals in partnership with the relevant Military Command for capability employment (eg. Joint Capabilities Group for C-UAS). 

Success would be measured against whether this reform closes the machine-speed accountability gap with human control in AI-versus-AI cyber engagements, establishing a bounded, Accountable Officer approved Control Model that allows defensive AI systems to respond within pre-authorised limits without diffusing responsibility for their actions. Central to this is operational rehearsals by the relevant ADF Military Command of the pre-authorisation framework. This will validate the effectiveness of the processes within conditions that simulate real-time AI-enabled threats. This option directly closes the governance gap identified in Problem Identification, and keeps accountability anchored to a named Accountable Officer rather than diffusing it, consistent with core principles and values of existing Defence policy.

Policy Recommendation

This paper proposes Option 1, amend Policy Settings for Responsible Use of Artificial Intelligence in Defence to include a new clause under Section 3: Proportionate Controls, to differentiate control model applications between AI-enabled systems with kinetic outcomes and AI-enabled systems with non-kinetic outcomes. This requirement would stipulate that any operation of an AI-enabled system to kinetically engage a target with irreversible outcomes must have positive human chain of command authorisation. Additionally, this control is not required for AI-enabled systems that do not result in an irreversible, non-kinetic outcome.

This option is the most achievable and effective out of the 3 models, as the amendment effectively addresses modernising the current policy to meet the operational realities of AI-enabled warfare. Notably, this allows greater diffusion and lower levels of control for non kinetic AI-enabled systems, while upholding Australia’s commitments to the 2023 Political Declaration. 

Governance and oversight of this amendment would remain anchored to the Accountable Officer already established in the Policy Settings for Responsible Use of Artificial Intelligence in Defence. The primary effect of this amendment to the Accountable Officer is that it would mandate different control model applications for AI-enabled systems depending on their outcomes. Operational rehearsals remain critical to ensuring policy alignment meets warfighting user requirements within the both kinetic and non-kinetic domains.

Funding requirements scale based on the extent of implementation led by DIAC. For the governance change to differentiate kinetic control models for AI-enabled systems, the costs would be relatively modest ($1 – 3 million) over a 1-2 year period. However, comprehensive operational rehearsals for Military Commanders and effective training for Accountable Officers could range from $5 – 15 million over a 1-2 year period depending on the scope of engagement activities. These figures are not inclusive of the cost of AI-enabled system acquisition.

Barriers and Risks

The key challenge to this policy amendment is likely to be whether it remains operationally relevant in a machine-speed cyberwar (Five Eyes security agencies, 2026) and if Military Commanders actually trust and use AI-enabled systems as intended (Horowitz et al, 2024). One of the biggest reasons for this is insufficient rehearsal of operational risk, particularly when it relates to irreversible kinetic outcomes. Accountable Officers and Operational Commanders may never be exposed to a realistic AI-enabled engagement before real-world application. A global study found that humans interact with AI systems based on their bias and education (Horowitz et al, 2024). Key findings conclude; ‘Those with the lowest levels of experience, knowledge, and/or familiarity (background) are relatively more averse to AI; people with middle levels of background are relatively over-reliant on AI, and those with the highest levels of background are relatively appropriately reliant on AI’ (Horowitz et al, 2024). In the context of military operations, Military Commanders may not fully understand what risks they are implicitly accepting when applying ‘positive human chain of command’ to kinetic effects if they are not appropriately educated and experienced with AI use.  

To frame this more simply: positive human command and control requires meaningful human understanding. This cannot be achieved solely through policy amendments. It must be de-risked through exposure to realistic AI-enabled engagements in training before real-world execution. This is why funding operational rehearsals for ADF Commanders is critical to success. The ADFs own exercise reporting illustrates the inherent need for test and trial of emerging capabilities in the field (Klinger, 2026). Implementation of this policy amendment through meaningful operational rehearsals will reduce the gap between policy compliance and operational readiness.

Secondly, from a governance and accountability perspective, ambiguity around 'kinetic' vs 'non-kinetic' in the context of AI is likely to blur traditional distinctions. Some commentators have argued that the consequence of this could be to default to higher levels of governance and control than what the policy intends (Kissell, 2026). This policy amendment mitigates this ambiguity by adopting a lens of reversible and irreversible outcomes. However, critics could argue there is still room for dispute over which control model applies given the interconnectedness of operational technology and information technology, and how technology proliferation is only increasing (Australian Signals Directorate, 2025). 

Finally, key allies (critically the Five Eyes alliance) may adopt different approaches to human control of their AI-enabled systems. This divergence could create interoperability challenges during coalition operations, particularly where there is a requirement to integrate AI-enabled sensor, decision support and targeting systems (Rogers, 2026). This could increase risk of the ADF operating AI systems outside of policy and legislative controls. Additionally, it could hinder operational effectiveness due to increased decision timelines resultant from complex policy and legal arrangements.

References

Australian Government Department of Defence 2026, 2026 National Defence Strategy, Australian Government, Canberra, viewed 20 September 2026, https://www.defence.gov.au/about/strategic-planning/2026-national-defence-strategy-2026-integrated-investment-program

Australian Government Department of Defence 2026, Policy Settings for Responsible Use of Artificial Intelligence in Defence, Australian Government, Canberra, viewed 20 September 2026, https://www.defence.gov.au/sites/default/files/2026-03/Policy-Settings-for-Responsible-Use-of-Artificial-Intelligence-in-Defence-%5BOFFICIAL%5D.pdf

Australian Government Department of Home Affairs 2023, 2023–2030 Australian Cyber Security Strategy, Australian Government, Canberra, viewed 20 September 2026, https://www.homeaffairs.gov.au/cyber-security-subsite/Pages/2023-2030-australian-cyber-security-strategy.aspx

Australian Signals Directorate (ASD) 2025, Annual Cyber Threat Report 2024–2025, Australian Government, viewed 20 September 2026, https://www.cyber.gov.au/about-us/view-all-content/reports-and-statistics/annual-cyber-threat-report-2024-2025

Boulanin, V, Davison, N, Goussac, N & Peldán Carlsson, M 2020, Limits on autonomy in weapon systems: identifying practical elements of human control, Stockholm International Peace Research Institute and International Committee of the Red Cross, Stockholm and Geneva, June, viewed 20 September 2026, https://www.icrc.org/sites/default/files/document/file_list/icrc_sipri_limits_on_autonomy_june_2020.pdf. 

Conroy, P (Minister for Defence Industry) 2026, 'Albanese Government to invest up to $7 billion in counter drone defence', media release, 21 April, viewed 20 September 2026, https://www.minister.defence.gov.au/media-releases/2026-04-21/albanese-government-invest-up-7-billion-counter-drone-defence. 

Ekelhof, M., 2019. Moving beyond semantics on autonomous weapons: Meaningful human control in operation. Global Policy, 10(3), pp.343-348.

Five Eyes cyber security agencies 2026, Five Eyes cyber security agencies statement, Australian Cyber Security Centre, Australian Signals Directorate, 22 June, viewed 20 September 2026, https://www.cyber.gov.au/about-us/view-all-content/news/five-eyes-cyber-security-agencies-statement. 

Horowitz, MC & Kahn, L 2024, 'Bending the automation bias curve: a study of human and AI-based decision making in national security contexts', International Studies Quarterly, vol. 68, no. 2, sqae020, doi:10.1093/isq/sqae020. 

ICRC (International Committee of the Red Cross) 2006, A guide to the legal review of new weapons, means and methods of warfare: measures to implement Article 36 of Additional Protocol I of 1977, ICRC, Geneva, viewed 20 September 2026, https://www.icrc.org/en/publication/0902-guide-legal-review-new-weapons-means-and-methods-warfare-measures-implement-article. 

Klingner, T 2026, 'Counter-drone tech put to the test', Department of Defence news, 5 August, viewed 20 September 2026, https://news.defence.gov.au/army/activity/2026-08-05/counter-drone-tech-put-test

Priyandita, G. 2026, ‘NDS 2026 – More resourcing for a complex cyber threat environment’, The Strategist, 12 May, viewed 20 September 2026, https://www.aspistrategist.org.au/nds-2026-more-resourcing-for-a-complex-cyber-threat-environment/

Protocol Additional to the Geneva Conventions of 12 August 1949, and relating to the Protection of Victims of International Armed Conflicts (Protocol I), opened for signature 8 June 1977, 1125 UNTS 3 (entered into force 7 December 1978) 

Rogers, M 2026, 'AUKUS partners test drone swarming tactics', Department of Defence news, 22 June, viewed 20 September 2026, https://www.defence.gov.au/news-events/news/2026-06-22/aukus-partners-test-drone-swarming-tactics. 

Santoni de Sio, F. and Mecacci, G., 2021. Four responsibility gaps with artificial intelligence: Why they matter and how to address them. Philosophy & technology, 34(4), pp.1057-1084.

Turillazzi, A. 2026, ‘The ADF is integrating AI faster than it can govern it’, The Strategist, 24 June, viewed 20 September 2026, https://www.aspistrategist.org.au/the-adf-is-integrating-ai-faster-than-it-can-govern-it/

U.S. Department of State 2023, Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy, U.S. Department of State, Washington, DC, viewed 20 September 2026, https://www.state.gov/bureau-of-arms-control-deterrence-and-stability/political-declaration-on-responsible-military-use-of-artificial-intelligence-and-autonomy

Whyte, C. 2020, ‘Problems of Poison: New Paradigms and 'Agreed' Competition in the Era of AI-Enabled Cyber Operations’, in Jančárková, T., Lindström, L., Signoretti, M., Tolga, I. and Visky, G. (eds), 20/20 Vision: The Next Decade: Proceedings of the 12th International Conference on Cyber Conflict (CyCon 2020), NATO Cooperative Cyber Defence Centre of Excellence, Tallinn, pp. 215–232, doi:10.23919/CyCon49761.2020.9131717

The views and opinions expressed by Global Voices Fellows do not necessarily reflect those of the organisation or its staff.