AI: Hope or Hazard for Peace?
New evidence shows AI’s promise for peace processes — but without steering, it could drive us toward peril.
The past two weeks have felt like whiplash.
At the UN General Assembly in New York, esteemed Nobel laureates and former heads of state signed onto an urgent call for “red lines” on artificial intelligence, warning of risks from pandemics to mass manipulation. The Global Call for AI Red Lines urges governments to agree by 2026 on enforceable “do-not-cross” limits. As Nobel laureate Maria Ressa put it: “Without AI safeguards, we may soon face epistemic chaos, engineered pandemics, and systematic human rights violations.”
At the very same time, new evidence from peacebuilders — along with a comprehensive global review — shows how AI can already widen participation, strengthen peace processes, and help societies navigate conflict without violence.
The juxtaposition is striking. On one hand, AI is helping amplify marginalized voices and support fragile dialogues. On the other, unchecked AI could destabilize societies and erode human rights. Both perspectives are true — and both demand urgent attention.
During these moments of seemingly binary choices — AI as savior or threat — I find the recent words of Taiwan’s former digital minister Audrey Tang helpful: “Many people feel that with the rapid advance of AI, our future is like a car with only a gas pedal and a brake. But we’ve overlooked the most important part: the steering wheel.” We don’t need to slam on the brakes or floor the gas. We need to learn to steer.
AI on the agenda, everywhere
Over the last few weeks, I’ve moved from Washington, DC to New York and now Albania — and at each stop, the role of AI has been at the center of conversation.
In DC, at the World Bank’s Coalition for Reform, representatives of 98 governments examined how the social contract is weakening, with public trust in governments falling while trust in chatbots grows.
Then, during UNGA week, I attended several events where technology was either the central focus or a cross-cutting theme. Search for Common Ground convened a forum on youth entrepreneurship and leadership and the International Youth Foundation (IYF), the Georgetown University Collaborative on Global Children’s Issues, the New Lines Institute, and Quest Alliance examined AI’s impact across generations. And at Tech Salon NYC, Linda Raftree convened practitioners to grapple with how to scale humanitarian uptake of AI while embedding ethical guardrails.
I write this now from Albania, where I’m attending a UNDP workshop on social cohesion. Here, the government recently appointed an AI “Minister of Public Procurement” who recently ‘addressed parliament’ — a gesture that raised eyebrows while signaling the opportunity of harnessing trust in AI to strengthen governance.
Across all of these discussions, the mood was consistent: a mix of excitement and trepidation, eagerness to experiment coupled with anxiety about ethics, trust and risk.
How AI helps peace processes
Leveraging AI in Peace Processes: A Framework for Digital Dialogues, a new paper by CMI–Martti Ahtisaari Peace Foundation advisor Felix Kufus and University of Birmingham Associate Professor Martin Waehlisch, is one of the clearest guides to date. It highlights three contributions AI can already make to peace processes:
Inclusion: lowering barriers for youth, women, and displaced people who are often excluded.
Sense-making: using natural language processing to synthesize thousands of inputs into clear patterns.
Scalability: enabling multilingual, geographically dispersed dialogues that extend well beyond physical negotiation tables.
As the authors put it: “AI’s greatest potential in peace processes lies in augmenting understanding through advanced analysis and sensemaking of stakeholder inputs… and expanding participation beyond those physically present at negotiation tables.”
This is augmentation, not automation. AI works best when it helps human facilitators listen better — not when it tries to replace them.
Different language, different answer
Yet even in this supportive role, risks remain. AI can amplify existing asymmetries in power and narrative. A 2025 study by Steinert & Kazenwadel, ‘How user language affects conflict fatality estimates in ChatGPT,’ found that when asked in Hebrew about Israeli airstrikes in Gaza, ChatGPT-3.5 reported 34% fewer fatalities on average than when asked in Arabic. A similar disparity was found between Turkish and Kurdish queries about violence.
This shows how AI reproduces the asymmetries of language and narrative embedded in its training data. That is why initiatives like Lisan, which builds open resources for Yemeni, Sudanese, and Libyan Arabic, and Mistral’s Saba model, trained on Middle Eastern data, are essential. Without smaller, locally grounded models, AI will continue to reflect global biases rather than local realities.
A set of tested tools
Across the field, a set of AI-enabled tools to strengthen deliberation and participation has already been tested:
These are not prototypes. They have been used in UN consultations, national dialogues, and grassroots initiatives. They show that AI can already widen participation and strengthen legitimacy in peace processes.
Global review highlights adoption deficit
The first ever International Panel on the Information Environment (IPIE) report on Artificial Intelligence and Peacebuilding, launched last week, takes a broader lens. The expert panel (C. Zelizer, F. Ogenga, L.Schirch, E. Tauchnitz, P. N. Howard, S. Valenzuela) reviewed more than 600 documents on AI in peacebuilding and governance. Amongst its conclusions is not that AI is unproven — quite the opposite. Instead, there is a challenge in uptake and adoption.
The report points to:
Infrastructure gaps in conflict zones - unreliable electricity, internet, and secure devices.
Global North dominance, with most tools designed in English and poorly adapted elsewhere.
Donor-driven pilots that showcase innovation but rarely become sustainable or scale.
Ethical frameworks without enforcement, leaving many deployments unaccountable.
In other words, the slow adoption isn’t a lack of evidence or possibility. It’s integration into institutions.
Despite the barriers, the IPIE and CMI reports point to similar priorities:
Invest in smaller, regional models that reflect local languages and contexts.
Embed AI into long-term peace strategies, not short-term experiments.
Build local capacity — facilitators, linguists, civic actors.
Make ethical safeguards enforceable, not optional.
On the frontier of this best practice is Build Up, which has recently designed and deployed a tailored digital consultation tool for Sudan called Mersal. Run primarily through automated WhatsApp conversations and supported by social media outreach, Mersal has already engaged thousands of Sudanese across political, social, and gender divides. Guided by a diverse Sudanese steering committee, it offers a model for how AI-enabled platforms can support inclusive dialogue even in the midst of war — not as a substitute for peace processes, but as a way to widen participation and sustain a shared vision for the future.
As Tang reminds us: “AI isn’t a weapon that creates division, but a tool that builds bridges. It helps us span our differences, find uncommon ground, and make democracy more resilient.”
Steering AI towards peace
The last two weeks of meetings — in DC, New York, and now Albania — have made the tension vivid: the promise of AI for peace is real, and so are the perils.
We now have enough evidence, enough use cases, and enough frameworks to move beyond scattered pilots. But scaling without guardrails risks deepening the very crises we hope to solve.
The challenge, then, is not invention but intention: to harness AI for inclusion and dialogue, while drawing firm red lines against its most dangerous uses.
Lena Slachmuijlder is Senior Advisor, Digital Peacebuilding at Search for Common Ground and Co-Chairs the Council on Tech and Social Cohesion.