Expanding AI Risk Explorer to cover geopolitical conflict risk

The AI Risk Explorer (AIRE) (https://www.airiskexplorer.com/) is an open intelligence hub that continuously monitors emerging, large-scale AI risks and curates related evidence (capability evaluations, benchmarks, incidents). It’s built to help decision-makers anticipate, understand, and manage AI risks.

Current coverage in AI Risk Explorer (AIRE) focuses primarily on AI-enabled bio threats, loss-of-control, cyber and manipulation risks. The aim of the project is to expand the coverage to include how AI might directly or indirectly contribute to creating geopolitical conflict. This project will extend AIRE by developing structured, evidence-linked resources that enable researchers and policymakers to navigate the full extent of risks with greater clarity. Example work packages include:

1. Creating AIRE-style risk explainers covering six potential pathways to conflict escalation, inspired by a recent RAND report (https://www.tandfonline.com/doi/full/10.1080/00963402.2025.2515793#abstract):

• AI disrupting the balance of power, creating incentives for aggression or rapid shifts in strategic advantage.

• Fear of adversary AI capabilities, leading states to consider preventive or preemptive actions.

• AI reducing the perceived cost of conflict, e.g., through automation, increased precision, or rapid decision cycles.

• AI contributing to societal instability, where internal disorder or disruption increases the probability of external conflict.

• AI systems initiating or triggering conflict on their own, whether through misinterpretation, faulty automation, or unintended behavior.

• AI influencing or degrading human decision-making, including overreliance, cognitive overload, or manipulated information environments.

1. Developing a taxonomy of risk drivers relevant to the pathways above

2. Creating risk assessments that measure risk levels and their evolution, based on the taxonomy of risk drivers and other relevant information

3. Curating databases of evidence (e.g., benchmarks, incidents, evaluations), informing analyses outlined above

4. Creating materials for public and targeted outreach (manuals, policy briefs, analyses, etc.)

Geopolitical conflict is both a major catastrophic AI risk area as well as an amplifier of other catastrophic AI risks. Creating a detailed breakdown of possible scenarios and curating a database of relevant resources advances AI safety by:

• Clarifying how AI could trigger or worsen major international crises.

• Identifying upstream signals that might reveal emerging instability.

• Supporting governance design by highlighting where oversight, transparency, and communication norms matter most.

• Providing concrete leverage points for policymakers and researchers to focus research and mitigation efforts on (e.g., confidence-building measures, communication protocols, etc).

Currently, we find that there is insufficient monitoring, especially ongoing and structured monitoring of AI-related geopolitical conflict risks, based on our survey of publicly available resources as well as discussions with policymakers.

Running a SPAR project on the risk AI-related geopolitical conflict also fits into the extended theory of change of the AI Risk Explorer, as outlined below:

The International Scientific Report on the Safety of Advanced AI, published in January 2025 highlights differing expert opinions on how quickly AI capabilities might advance. This uncertainty creates what the authors call an “evidence dilemma”: Regulators must choose between acting early based on limited evidence or risking being overtaken by sudden leaps in AI capabilities and the resulting hazards.

The “evidence dilemma” is exacerbated by a dire shortage of visibility into AI safety research, which hinders the management of risks of artificial intelligence globally. The Singapore Consensus on Global AI Safety Research Priorities - a major report on AI risk management - highlights a consensus of international experts, indicating that AI risk assessment is one of the three major research areas in need of the greatest development.

To improve awareness among policy- and decision-makers, we launched AI Risk Explorer (AIRE) (https://www.airiskexplorer.com/) in October 2025. AIRE is an open intelligence hub that continuously monitors emerging, large-scale AI risks and curates related evidence (capability evaluations, benchmarks, incidents). It’s built to help decision-makers anticipate, understand, and manage AI risks based on open-source intelligence (OSINT).

OSINT is widely used in a variety of fields, including law enforcement, national security, public health, cybersecurity, and business risk management. Advantages of OSINT include easily accessible evidence base for transparent, accountable decision-making; ability to generate timely, broad coverage of risks; ability to generate shared situational awareness among many actors (easy sharing of intelligence); and in many cases, lower costs than specialised, classified intelligence.

In 2026, we plan to expand AI Risk Explorer to cover new risks and provide new instruments for the analysis of systemic AI risks. As of now, we also actively work with policymakers (mostly, but not limited to, the EU area) to help them understand emerging AI risks with AIRE. Our further plans include building a community of experts around AIRE to increase its reach and impact even more. Running a project during SPAR would help us achieve all of the aforementioned goals and directly contribute to the betterment of AI risk management worldwide.

Mentees will act as Risk Analysts in areas such as:

Risk Taxonomy Analysis and Creation: Mentees will investigate the potential pathways to AI-related geopolitical conflict and create a taxonomy of related risk drivers, complete with risk thresholds related to specific drivers.

Risk Assessment: Mentees will collect evidence to assess the current level of geopolitical conflict risk and its evolution in time, per the taxonomy above.

Scenario Design: Mentees will draft hypothetical "Threat Scenarios" to serve as examples in respective explainers.

Resource Aggregation: Mentees will investigate safety literature and other sources to find evaluations, incidents and benchmarks to populate the AIRE database.

Public outreach: Mentees will create materials for public and targeted outreach (manuals, policy briefs, analyses, etc.)

• previous research experience related to geopolitical conflict, role of AI in geopolitical conflict, or other systemic risks from AI

• ability to collect and summarize findings from multiple sources (scientific papers, reports, others)

• ability to write clearly in technical, yet accessible language

Ability to attend meetings within business hours in European time zones is preferred.