With the massive push toward an artificial intelligence (AI)-centric future, we explore the implications of AI on important areas of peace research, particularly related to AI’s effects on democracy, mass surveillance, and warfare. In our informing practice section, we note that the documented harms and potential opportunities of AI are not equally balanced. Rather, AI poses far greater harm to peace—especially for marginalized and racialized communities—than it helps to create a more peaceful and just future, without necessary regulation, transparency, and accountability.
What We’re Reading
For each of the articles mentioned below, we include the central research question, the research methodology used, and the authors’ main findings.
Digital Identities: ‘Proactive’ Policing of Bodies Beyond Borders
Citation: Mahmoudi, M., & Willis, G. D. (2025). From ‘ID’ to proactive profiling: Identity, selective governance and the speculative making of bodies as borders. Security Dialogue, 56(4), 438-458.
How are digital identities that transnational technology corporations commodify and provide to governments used to surveil (vulnerable) populations far beyond physical borders?
Mahmoudi and Willis conducted qualitative and ethnographic research over nine years—spanning Latin America, Europe, and the United States—where they observed, interviewed, and analyzed documents related to refugees and family members of missing persons. They studied how states and private corporations use digital identity (ID) verification systems and predictive AI technologies to preemptively profile, police, and exclude marginalized populations at and beyond border checkpoints.
- Biometrics link identities to physical bodies, whereas digital IDs are metadata about online behaviors and social relations used to predict an individual’s future intent.
- Digital identity systems have allowed states and private corporations to govern borders speculatively, based on online metadata gathered from individuals long before they reach a physical border.
- The digital ID ecosystem relies heavily on data that are harvested from marginalized, displaced, and undocumented populations; transnational technology corporations exploit this data and sell their surveillance platforms and services to states for (beyond) border securitization—such as seen by Microsoft’s partnership with ICE.
- AI technology and surveillance are built and sustained on unjust foundations, requiring unequal social class relations and racial capitalism to function—to extract value, police for profit, and deny entry to globally marginalized populations into wealthy countries.
How does the U.S. public respond to military use of AI in intelligence gathering and targeted bombings?
Citation: Zwald, Z., Kennedy, R., & Ozer, A. (2025). The political viability of AI on the battlefield: Examining US public support, trust, and blame dynamics. Journal of Peace Research, 62(6), 1750-1764.
The U.S. military is developing artificial intelligence (AI) tools to enhance intelligence, surveillance, and reconnaissance (ISR) capabilities and autonomous weapons systems (AWS). The level of U.S. public support for the military’s use of AI can either legitimize greater machine autonomy—if the public expresses a higher level of trust in AI systems—or constrain it—if the public expresses low levels of trust in AI systems. The authors conducted “three vignette-based survey experiments” with around 800 U.S.-based respondents to understand the public’s trust in the military’s use of AI.
- They find that (1) the U.S. public’s support for AI in military operations is “heavily context-specific,” (2) the degree to which there is human control or fully machine autonomous systems affects the “likelihood of trust in reliability,” and (3) “the public is less likely to blame the human operator [when a military strike is] partially and fully autonomous…relative to when the human controls both identifying and striking the target.”
- The first two vignettes focus on “the reliability of AI-generated intelligence analysis,” measuring responses to different scenarios that vary by the “decision-making autonomy of the team identifying an enemy target (human-only versus human-machine versus machine-only)” and whether there is agreement between human-based and machine-based intelligence.
- The first two experiments reveal no significant difference in public support on whether a human intelligence officer or fully autonomous AI provided the intelligence but, “the public is more likely to support a strike when the human and machine both make a positive identification.”
- The third vignette tested the public’s response to a drone strike on an enemy target and measured levels of trust “regarding human-machine decision-making autonomy.” This experiment had two phases wherein the second phase reveals that the strike hit not an enemy target but instead civilians, killing seven. The first phase of this experiment revealed “no significant evidence that the drone’s level of autonomy affects the public’s trust” that a target was correctly identified. The second phase of the experiment revealed that “increasing machine autonomy decreases how much the public blames the soldier,” for the loss of innocent lives and the public becomes more critical about the use of drones in warfare.
- One critical insight from this paper is that if human drone operators understand that the public will blame them less for strikes that kill civilians when AI is more involved in the decision-making process, then they are incentivized to “cede decision-making autonomy to the machine…as a way to insulate themselves from blame and accountability.”
AI and Democracy
Citation: Summerfield, C., Argyle, L.P., Bakker, M., et al. (2025). The impact of advanced AI systems on democracy. Nature Human Behaviour 9(12), 2420–2430. https://doi.org/10.1038/s41562-025-02309-z
In this review essay—which synthesizes existing evidence rather than generating new data—23 researchers from universities, civil society organizations, and the companies building AI models map the effects of advanced AI on democratic life.
- The authors categorize the effects into three types:
- Epistemic impacts change what citizens can know and decide.
- Material impacts affect the machinery of elections and governance.
- Foundational impacts affect the principles that hold democratic systems together, including accountability and the distribution of power.
- Each category contains both harm and opportunity; the outcome depends on choices about how AI systems and democratic institutions are designed to work together.
- AI systems can shift political attitudes in experiments, including on questions where people initially disagree. Whether they depolarize (or polarize) populations at scale remains unresolved.
- The most concrete potential damage of AI is material. Voter roll purges, mass registration challenges filed faster than officials can review them, phishing attacks on electoral bodies, and synthetic audio aimed at suppressing turnout are already in use. These require no breakthrough in technology and cost almost nothing for those wishing to wreak havoc.
- The authors identify gendered attacks as a distinct mechanism. Generative tools are being used to produce sexual deepfakes and defamation aimed at women in politics and activism, with the aim of driving them out of public life. The authors cite reporting that women in U.S. politics receive far more online abuse than their male counterparts.
- There is a small but real evidence base for AI’s positive role in peacebuilding and democratic deliberation. AI systems have been used to help groups converge on collective statements preferred to human-drafted ones, to increase the share of ideas contributed by women in mixed-gender political discussions among Afghan citizens, and, in one 2025 study, to find common ground between Israeli and Palestinian peacebuilders.
Informing Practice
Artificial intelligence dominates headlines and has become part of everyday life, even to those who try to minimize their use of it. Every web search produces an AI summary unless the user takes active steps to turn that function off. Hyperscalers—the giant cloud providers like Amazon and Google—are building data centers in nearly all 50 states with thousands more planned, often amid local community opposition due to the centers’ extraordinary electricity and water use. Global investment in AI is forecasted to exceed $1 trillion in 2026 and accounts for 45% of the S&P 500 total mark cap, now dominating the U.S. equities and credit market. There are prevalent fears about AI destabilizing job markets, outwitting human controls, or even “destroy[ing] humans”.
The articles reviewed in this month’s round-up examine the effects of AI on important questions in the peace and security field: how is AI applied in mass surveillance and contributing to militarization of immigration? Can the public meaningful constrain militaries’ use of AI in intelligence operations or in the development of autonomous weapons systems? How is AI affecting democracies?
From this snapshot of peace research, we observe that the documented harms and opportunities for AI are not equally balanced. Without greater accountability, transparency, and regulation of AI, the harms far outweigh the opportunities. Likewise, the burden and the benefit of AI fall along predictable social, economic, and racial fractures. The very wealthy are posed to accumulate even more wealth, while marginalized, racialized communities are subjected to a higher degree of scrutiny, mass surveillance, state violence, and militarization.
Harms
- Mahmoudi and Willis (2025) highlights how the world’s largest tech companies provide the necessary data and AI platforms for governments’ mass surveillance and border securitization operations. Investigative reporting from The Guardian and others revealed the Immigrations and Customs Enforcement (ICE) tripled the amount of data stored on Microsoft’s Azure cloud platform in the six months leading up to January 2026. ICE appears to be using Microsoft’s AI-driven tools to search and analyze data stored on the cloud. And the benefit is cross-cutting—Microsoft has been awarded tens of millions of dollars in contracts as the ICE budget has ballooned under the Trump Administration. Microsoft denies that its platforms are used for mass surveillance of citizens while ICE has not revealed what kind of data or analyses it conducts using AI.
- Zwald, Kennedy, and Ozer (2025) note the reality that militaries are applying AI to the battlefield. In the first 24 hours of the U.S. war in Iran, the AI-powered Maven Smart System created by Palantir, “generated hundreds of strike coordinates…enabling the U.S. to hit 1,000 targets,” including Shajareh Tayyebeh elementary school in Minab which killed 168 people, most of them under the age of 12. Congress has questioned the extent to which there was human control over the AI-enabled weapons systems involved without a clear answer from military leaders. Not only does this raise serious concerns regarding transparency and accountability—especially considering the Zwald, Kennedy, and Ozer (2025) finding that drone operators are incentivized to defer to machines precisely to avoid accountability—but amplifies the tragedy and inhumanity of the strike when considering how avoidable and irresponsible the U.S. war effort is in the first place.
- Summerfield et al (2025) find that the most concrete damage to democracy from AI is from administrative disruption. ‘Voter fraud’ activists could use AI to purge voters rolls and overwhelm elections officials—as experts raised in 2024 with conservative activists in the U.S. attending trainings on EagleAI, a platform with filled with nation-wide voter roll data. Or, AI-generated deepfakes can easily and rapidly spread misinformation and propaganda surrounding elections, like a fake Joe Biden robocall telling voters to sit out of upcoming elections. As AI tools become more sophisticated and better funded over time, the implications for election integrity become more severe.
Opportunities
- Summerfield and colleagues (2025) also documented positive applications of AI in peacebuilding. AI systems helped groups reach agreements they preferred to human-written ones, increased women’s contributions to mixed-gender political discussions in Afghanistan, and helped Israeli and Palestinian peacebuilders find common ground. Mediators Beyond Borders International (MBBI) has produced guidelines stating that AI should never speak on behalf of a community in place of its members and that the environmental cost should be taken into account when deciding whether to use these tools.
The guidelines produced by Mediators Beyond Borders International point to an incredible need as society comes to manage—or not—the AI technological revolution. Namely, the potential of this technology to more quickly process data that deepens inequality and kills must be actively prevented and managed. Both Mahmoudi and Willis (2025) and Summerfield et al (2025) argue that AI rests on unjust foundations. It is only as accurate and unbiased as the training data that feeds these systems. AI systems that are fed out-of-date information (as suggested in reporting on the bombing on the elementary school in Iran) or data that is incredibly biased towards identifying “threats” based on an individual’s identity markers only serve to amplify existing violent and unjust systems on a global scale. The natural resources that are exploited to feed this vicious cycle cannot be ignored either, especially in an era of climate change and rapid biodiversity loss, further extenuating the harms of AI beyond our current generation.
Greater regulation, accountability, and transparency is needed—but it also raises deeper questions about how society should function. AI is a tool that can reflect our reality back to us, and perhaps we should think about the necessary reforms needed to make our society more just, equitable, and peaceful regardless of what era of technological innovation defines the present-day.
Further Reading:
On AI in conflict resolution
Institute for Integrated Transitions. (2026, August 13). AI conflict resolution advice has improved but is still failing, new IFIT study finds. https://ifit-transitions.org/blog/ai-conflict-resolution-advice-has-improved-but-is-still-failing-new-ifit-study-finds/
Moshtagi, R., Cortez, J., & Sohn, A. (2025, May 27). AI and the future of conflict resolution: How can artificial intelligence improve peace negotiations? Belfer Center for Science and International Affairs, Harvard Kennedy School. https://www.belfercenter.org/research-analysis/ai-and-future-conflict-resolution-how-can-artificial-intelligence-improve-peace
On AI and Elections Integrity
https://carnegieendowment.org/research/2024/12/can-democracy-survive-the-disruptive-power-of-ai
https://www.brennancenter.org/our-work/analysis-opinion/election-year-risks-ai
Photo credit: Public Domain Pictures
