This paper describes how AI can theoretically contribute to counterterrorism operations, as well as some areas where it is already used. It presents some of the problems with the current use of AI for these purposes, examining both practical limitations and implications for human rights. It also considers the opportunities and risks posed by increased use of AI as part of counterterrorism activity by governments. It makes the case that use of AI in counterterrorism is not inherently wrong, and suggests some necessary conditions for legitimate use of AI as part of a predictive approach to counterterrorism on the part of liberal democratic states.

Highlights:

• The use of predictive artificial intelligence (AI) in countering terrorism is often assumed to have a deleterious effect on human rights, generating spectres of ‘pre-crime’ punishment and surveillance states. However, the well-regulated use of new capabilities may enhance states’ abilities to protect citizens’ right to life, while at the same time improving adherence to principles intended to protect other human rights, such as transparency, proportionality and freedom from unfair discrimination. The same regulatory framework could also contribute to safeguarding against broader misuse of related technologies.
• Most states focus on preventing terrorist attacks, rather than reacting to them. As such, prediction is already central to effective counterterrorism. AI allows higher volumes of data to be analysed, and may perceive patterns in those data that would, for reasons of both volume and dimensionality, otherwise be beyond the capacity of human interpretation. The impact of this is that traditional methods of investigation that work outwards from known suspects may be supplemented by methods that analyse the activity of a broad section of an entire population to identify previously unknown threats.
• Developments in AI have amplified the ability to conduct surveillance without being constrained by resources. Facial recognition technology, for instance, may enable the complete automation of surveillance using CCTV in public places in the near future.
• The current way predictive AI capabilities are used presents a number of interrelated problems from both a human rights and a practical perspective. Where limitations and regulations do exist, they may have the effect of curtailing the utility of approaches that apply AI, while not necessarily safeguarding human rights to an adequate extent.
• The infringement of privacy associated with the automated analysis of certain types of public data is not wrong in principle, but the analysis must be conducted within a robust legal and policy framework that places sensible limitations on interventions based on its results.
• In future, broader access to less intrusive aspects of public data, direct regulation of how those data are used – including oversight of activities by private-sector actors – and the imposition of technical as well as regulatory safeguards may improve both operational performance and compliance with human rights legislation. It is important that any such measures proceed in a manner that is sensitive to the impact on other rights such as freedom of expression, and freedom of association and assembly.