By: Helen Toner, John Bansemer, Kyle Crichton, Matthew Burtell, Thomas Woodside, Anat Lior, Andrew Lohn, Ashwin Acharya, Beba Cibralic, Chris Painter, Cullen O’Keefe, Iason Gabriel, Kathleen Fisher, Ketan Ramakrishnan, Krystal Jackson, Noam Kolt, Rebecca Crootof, and Samrat Chatterjee

The concept of artificial intelligence systems that actively pursue goals—known as AI “agents”—is not new. But over the last year or two, progress in large language models (LLMs) has sparked a wave of excitement among AI developers about the possibility of creating sophisticated, general-purpose AI agents in the near future. Startups and major technology companies have announced their intent to build and sell AI agents that can act as personal assistants, virtual employees, software engineers, and more. While current systems remain somewhat rudimentary, they are improving quickly. Widespread deployment of highly capable AI agents could have transformative effects on society and the economy. This workshop report describes findings from a recent CSET-led workshop on the policy implications of increasingly “agentic” AI systems.

In the absence of a consensus definition of an “agent,” we describe four characteristics of increasingly agentic AI systems: they pursue more complex goals in more complex environments, exhibiting independent planning and adaptation to directly take actions in virtual or real-world environments. These characteristics help to establish how, for example, a cyber-offense agent that could autonomously carry out a cyber intrusion would be more agentic than a chatbot advising a human hacker. A “CEO-AI” that could run a company without human intervention would likewise be more agentic than an AI acting as a personal assistant.

At present, general-purpose LLM-based agents are the subject of significant interest among AI developers and investors. These agents consist of an advanced LLM (or multimodal model) that uses “scaffolding” software to interface with external environments and tools such as a browser or code interpreter. Proof-of-concept products that can, for example, write code, order food deliveries, and help manage customer relationships are already on the market, and many relevant players believe that the coming years will see rapid progress.

In addition to the many potential benefits that AI agents will likely bring, they may also exacerbate a range of existing AI-related issues and even create new challenges. The ability of agents to pursue complex goals without human intervention could lead to more serious accidents; facilitate misuse by scammers, cybercriminals, and others; and create new challenges in allocating responsibility when harms materialize. Existing data governance and privacy issues may be heightened by developers’ interest in using data to create agents that can be tailored to a specific user or context. If highly capable agents reach widespread use, users may become vulnerable to skill fade and dependency, agents may collude with one another in undesirable ways, and significant labor impacts could materialize as an increasing range of currently human-performed tasks become automated.

To manage these challenges, our workshop participants discussed three categories of interventions:

• Measurement and evaluation: At present, our ability to assess the capabilities and real-world impacts of AI agents is very limited. Developing better methodologies to track improvements in the capabilities of AI agents themselves, and to collect ecological data about their impacts on the world, would make it more feasible to anticipate and adapt to future progress.

• Technical guardrails: Governance objectives such as visibility, control, trustworthiness, as well as security and privacy can be supported by the thoughtful design of AI agents and the technical ecosystems around them. However, there may be trade-offs between different objectives. For example, many mechanisms that would promote visibility into and control over the operations of AI agents may be in tension with design choices that would prioritize privacy and security.

• Legal guardrails: Many existing areas of law—including agency law, corporate law, contract law, criminal law, tort law, property law, and insurance law—will play a role in how the impacts of AI agents are managed. Areas where contention may arise when attempting to apply existing legal doctrines include questions about the “state of mind” of AI agents, the legal personhood of AI agents, how industry standards could be used to evaluate negligence, and how existing principal-agent frameworks should apply in situations involving AI agents.

While it is far from clear how AI agents will develop, the level of interest and investment in this technology from AI developers means that policymakers should understand the potential implications and intervention points. For now, valuable steps could include improving measurement and evaluation of AI agents’ capabilities and impacts, deeper consideration of how technical guardrails can support multiple governance objectives, and analysis of how existing legal doctrines may need to be adjusted or updated to handle more sophisticated AI agents.

Center for Security and Emerging Technology, October 2024

By: Yasmin Afina and Sarah Grand-Clément

Data plays a critical role in the training, testing and use of artificial intelligence (AI), including in the military domain. Research and development for AI-enabled military solutions is proceeding at breakneck speed, and the important role data plays in shaping these technologies has implications and, at times, raises concerns. These issues are increasingly subject to scrutiny and range from difficulty in finding or creating training and testing data relevant to the military domain, to (harmful) biases in training data sets, as well as their susceptibility to cyberattacks and interference (for example, data poisoning). Yet pathways and governance solutions to address these issues remain scarce and very much underexplored.

This paper aims to fill this gap by first providing a comprehensive overview on data issues surrounding the development, deployment and use of AI. It then explores data governance practices from civilian applications to identify lessons for military applications, as well as highlight any limitations to such an approach. The paper concludes with an overview of possible policy and governance approaches to data practices surrounding military AI to foster the responsible development, testing, deployment and use of AI in the military domain.

CIGI Papers No. 308 — October 2024

By: Jennifer Tang, Tiffany Saade, and Steve Kelly

Cutting-edge advances in artificial intelligence (AI) are taking the world by storm, driven by a massive surge of investment, countless new start-ups, and regular technological breakthroughs. AI presents key opportunities within cybersecurity, but concerns remain regarding the ways malicious actors might also use the technology. In this study, the Institute for Security and Technology (IST) seeks to paint a comprehensive picture of the state of play— cutting through vagaries and product marketing hype, providing our outlook for the near future, and most importantly, suggesting ways in which the case for optimism can be realized.

The report concludes that in the near term, AI offers a significant advantage to cyber defenders, particularly those who can capitalize on their "home field" advantage and firstmover status. However, sophisticated threat actors are also leveraging AI to enhance their capabilities, making continued investment and innovation in AI-enabled cyber defense crucial. At this time of writing, AI is not yet unlocking novel capabilities or outcomes, but instead represents a significant leap in speed, scale, and completeness.

This work is the foundation of a broader IST project to better understand which areas of cybersecurity require the greatest collective focus and alignment—for example, greater opportunities for accelerating threat intelligence collection and response, democratized tools for automating defenses, and/or developing the means for scaling security across disparate platforms—and to design a set of actionable technical and policy recommendations in pursuit of a secure, sustainable digital ecosystem.

The Institute for Security and Technology, October 2024

By Ramesh Subramanian, Eddan Katz

The Internet has been integral to the globalization of a range of goods and production, from intellectual property and scientific research to political discourse and cultural symbols. Yet the ease with which it allows information to flow at a global level presents enormous regulatory challenges. Understanding if, when, and how the law should regulate online, international flows of information requires a firm grasp of past, present, and future patterns of information flow, and their political, economic, social, and cultural consequences.

In The Global Flow of Information, specialists from law, economics, public policy, international studies, and other disciplines probe the issues that lie at the intersection of globalization, law, and technology, and pay particular attention to the wider contextual question of Internet regulation in a globalized world. While individual essays examine everything from the pharmaceutical industry to television to “information warfare” against suspected enemies of the state, all contributors address the fundamental question of whether or not the flow of information across national borders can be controlled, and what role the law should play in regulating global information flows.

New York: NYU Press, 2011.

By Nina Jankowicz, Isabella Gomez-O’Keefe, Lauren Hoffman and Andrea Vidal Becker

Technology facilitated gender-based violence (TFGBV) is not an in­tractable problem. But it must no longer be the responsibility solely of women’s advocacy groups. Others – technology companies, governments, civil society organizations, law enforcement, businesses, schools – must step up and work in unison to combat TFGBV in order to to reflect its main­streamed effects on society. This report, drawing on a case study around the online harassment of Australian eSafety Commissioner Julie Inman Grant, assesses the state of research on TFGBV as well as recent global policy progress made on this issue, and offers a number of practical solutions to make women and girls safer online. The authors argue that TFGBV must be mainstreamed to be mitigated, centering women’s experiences in broader policy debates. Technology companies, governments, civic tech organizations, law enforcement, employers, schools, and others must mainstream their work to combat TFGBV to reflect its mainstreamed effects on society. To this end, the authors recommend a number of practical solutions to the specific and pressing issues that women and girls face online today. Addressing the urgent changes described here will not only make women and girls safer and ensure their voices are heard, but also improve the safety and free expression for everyone who uses the internet, building more robust, representative democracies.

The recommendations are presented under the following themes:

• Ensuring platform accountability and action

• Urgently addressing deepfake image-based sexual abuse

• Supporting victims and survivors of TFGBV

• Deepening research and mainstreaming advocacy.

New York: Columbia University, Institute of Global Politics, 2024. 41p.

By Jai Gupta

This article aims to explore the relationship between weather and crime by answering the question: To what extent does the weather in cities with varying year-round temperatures have different relationships with both property and violent crime? The current literature has found a significant correlation between the increase in outside temperature with the increase in crime. However, this relationship has not been compared between warm and cold cities with relatively similar city variables, such as demographics, poverty rates, and others. The researcher performed a quantitative, correlational, ex post facto study to address this gap in the literature. The researcher examined three weather variables: average average temperature, average high temperature, and average low temperature across six cities. These six cities were split into three pairs, each pair having a warm and cold city. The researcher found four principle findings: cities with colder year-round temperatures had a more significant correlation between weather and crime; of the weather variables examined, the average low temperature had the strongest relationship with crimes across all cities; between violent and property crime, property crime had a stronger correlation with all weather variables on average; and lastly, as temperature increases across all weather variables, crime, on average, also increased. While more research should be conducted, these findings hope to serve local policy makers and law enforcement on better predicting crime.

Unpublished paper,  (April 30, 2024).

By  Aika Hui , Paul Salkovskis, Joshua Rumble-Browne

The current systematic review aimed to critically examine the growing body of literature proposing that there is an ‘intergenerational cycle’ of violence, whereby victims of abuse during childhood are posited to have a higher propensity of becoming perpetrators during adulthood. Specifically, this review examined whether there is quality evidence supporting the relationship between childhood sexual abuse victimisation and interpersonal violence perpetration (sexual/physical) in adult men. 20 studies published between 1992 and 2022 were included in this review. The quality of studies was systematically assessed to provide a weighted conclusion to the primary research question. Overall, there was limited evidence to confidently support or reject the link between childhood sexual abuse and physical and/or sexual violence in adulthood for men. The current review found that whilst there were studies that found associations on a univariate and multivariate level between childhood sexual abuse and interpersonal violence in adult men, only 25 % of the papers included in this review were deemed high-quality and significant methodological issues limit the validity of conclusions made. Addi tionally, a sizeable proportion of high-quality studies at both univariate and multivariate levels of analysis suggest mixed results or did not find a significant relationship. Implications for future research studies in this area in terms of the encouragement of a more critical stance towards the assumption of ‘intergenerational cycles’ of violence as well as recommendations for methodological improvements of studies were discussed.  

Aggression and Violent Behavior Volume 78, September–October 2024, 101928

By David Weisburd, Taryn Zastrow  , Kiseong Kuen  , Martin A. Andresen

Our paper reports on a systematic review of crime concentration studies over the last 35 years. We identify 47 papers that report on crime concentrations at a micro geographic unit of analysis. These papers produced 49 estimates of general crime concentration for crime incidents at streets segments for a specific cumulative proportion of crime. The median concentration for these estimates is 50 % of crime found at 4.5 % of streets, and 25 % of crime at 1.25 % of streets. The bandwidth for 50 % crime concentration is 9.3 % (1.7–11.0 %), and for 25 % crime concentration 3.0 % (0.4–3.4 %). Using the interquartile range to exclude outliers, we found a bandwidth of 2.5 % (3.2–5.7 %) for 50 % of crime, and 1.4 % (0.8–2.2 %) for 25 % of crime. Crime concentration was generally stronger for specific types of crime, and bandwidths of concentration were generally smaller. Using alternative measures of crime and different micro geographies produces similar conclusions regarding strong crime concentration at place with narrow bandwidths. Our review shows that a general framework of a law of crime concentration applies across a large number of cities across many regions around the world. 

Aggression and Violent Behavior Volume 78, September–October 2024, 101979

By Alex Armand, Paul Atwell, Joseph Gomes, Giuseppe Musillo, Yannik Schenk

This paper investigates the role of mass media in shaping racial tolerance and advancing civil rights in the post-WWII United States. We study the first attempt in the history of mass media to use a radio broadcast targeted at children to promote an inclusive American society. In 1946, amid persistent racial divisions, the popular radio series The Adventures of Superman launched Operation Intolerance, a sequence of new episodes promoting equality, rejecting racial discrimination, and exposing the KKK's bigotry. Using digitized historical data on U.S. radio stations and state-of-the-art radio propagation models, we compute geographic exposure to the broadcasts. Exploiting exogenous exposure to the broadcasts, we employ a cohort study design to analyze individual-level data from 1964 to 1980–a crucial period for civil rights activism and legislation in the United States. We find lasting impacts on those exposed as children, including increased support for civil rights, improved interracial relations, and more progressive political attitudes. These effects translate into greater alignment with the Civil Rights Movement, evidenced by increased support for protests and diminished institutional trust, and further manifested by reduced participation in the Vietnam War. Additionally, county-level panel data illustrate how areas covered by the broadcast in 1946 evolve towards less segregationist attitudes, a lower presence of the KKK, and an increase in civil rights activism and prominence in discourse.

Bonn: Institute of Labor Economics, 2024.

By The  Institute for Strategic Dialogue

The rapidly evolving online extremist landscape in New Zealand means new strategies for intervention are needed. This playbook – developed in consultation with New Zealand’s rich tapestry of civil society, communities, and practitioners engaged in prevention – offers practical frameworks for projects promoting positive online interventions to tackle online extremism. Based on an analysis of the rapidly evolving landscape of online extremism, the playbook takes stock of established and emerging intervention models. It brings together domestic and international best practices and suggests potential avenues for new positive intervention approaches. Finally, it reflects on practical considerations for programming, including monitoring and evaluation, safeguarding, operational security, and ethical considerations. This playbook examines the shift from violent groups to online extremism, highlighting digital literacy, audience communication, and proactive engagement with at-risk individuals.

London: Institute for Strategic Dialogue, 2024. 37p.

Edited by Henrietta Wilson, Olamie Samuel, and Dan Plesch

How did a journalist find out who was responsible for bombing hospitals in Syria from his desk in New York? How can South Sudanese monitors safely track and detail the weapons in their communities and make sure that global audiences take notice? How do researchers in London coordinate worldwide work uncovering global corruption? What are policymakers, lawyers, and intelligence agencies doing to keep up with and make use of these activities? In the age of Google, threats to human security are being tracked in completely new ways. Human rights abuses, political violence, nuclear weapons, corruption, radicalization, and conflict are all being monitored, analyzed, and documented. Although open source investigations are neither easy to conduct nor straightforward to apply, with diligence and effort, societies, agencies, and individuals have the potential to use them to strengthen security. This interdisciplinary book presents 18 original chapters by prize-winning practitioners, experts, and rising stars, detailing what open source investigations are and how they are carried out, and examining the opportunities and challenges they present to global transparency, accountability, and justice. It is essential reading for current and future digital investigators, journalists, and scholars of global governance, international relations, and humanitarian law, as well as anyone interested in the possibilities and dangers of this new field.

Singapore: World Scientific Publishing Europe Ltd, 2024.