AI and National Security: Scenarios Workshop Summary Report

Promote widespread diffusion and adoption of reliable AI tools to advance productivity and prosperity, while preventing adoption of unreliable AI in critical systems.

Invest in widespread skills development and broad access to data and computing power.

Strengthen measures to support workers displaced by AI-enabled automation, including retraining, income support, and AI-enabled learning tools.

Strengthen information-sharing protocols to facilitate research cooperation without unintended information flows.

Expand international cooperation and research on AI safety and security, including risks, technical safety, and governance mechanisms.

Build shared international understanding about global-scale AI risks and mitigations, including with rival powers, as a foundation for future cooperation if needed.

Limit use of AI systems in key military decisions and strengthen national and international rules on the use of autonomous weapons.

Strengthen national and international AI incident monitoring and build protocols for coordinated emergency response.

Develop redundant physical and digital repositories of information for knowledge preservation in the event of an accidental or necessary internet shutdown.

Implement transparency requirements and whistleblower protections for disclosures about dangerous AI systems or practices.

Harden digital systems from cyberattacks, limit access to bio-synthesis tools, and improve detection and response to novel pathogens — essential at all levels of AI development.

Promote rapid AI adoption by removing regulatory impediments, while strengthening incentives for the development and deployment of reliable AI systems and tools.

Prepare for the possibility of moderate-to-high job losses and economic disruption by enhancing policy measures for retraining, income support, business development, and cost-of-living reduction.

Monitor and mitigate AI adoption risks in key domains such as nuclear missile launch, critical infrastructure, and personal companionship, to avoid critical failures.

Prepare for the possibility of extreme job losses and economic disruption, including by developing new income distribution infrastructure.

Develop international systems to address severely widening income inequalities among countries driven by uneven AI adoption.

Establish joint international limits on AI adoption in key domains such as nuclear missile launch and critical infrastructure to avoid critical failures.

Establish joint international limits on the overall level of AI adoption and societal automation, if needed, to prevent excessive displacement of humans in decision making.

Build trust and defuse tension between rival powers to limit a military AI race that could result in rapid, unintended escalation and global conflict.

Limit use of AI systems in key military decisions and develop protocols for scaled and predictable sabotage and retaliation to reduce the risk of sudden destabilizing shifts in the balance of power.

Implement a robust international AI transparency and monitoring regime to provide timely alerts of emerging global AI risks, such as from imminent breakthroughs in AI capabilities.

Ensure that national and coordinated emergency response protocols are adequate to handle risks posed by the potential sudden emergence of much more capable AI systems.

Develop mechanisms to coordinate use of ASIs across countries to avoid ASI-enabled conflict among multiple controllable ASI powers.

Develop national or international mechanisms to prevent any single faction from gaining control of an ASI.

Develop technical verification and compliance measures to prevent malicious use and weaponization of ASI systems.

Prepare an international agreement with the ability to enforce a moratorium on ASI development if required.

Prepare an international agreement with mechanisms to coordinate use of ASIs across countries if safe ASI cooperation is seen as possible.

Adopt antitrust measures to prevent monopoly and singleton control of ASI in the market or politically.

Implement international agreement and cooperation to ensure inclusive and legitimate decision making around the development and use of ASI.

Develop the ability, via international cooperation if necessary, to prevent creation of ASI anywhere in the world until there is sufficient evidence that it can be controlled and aligned.

Prepare emergency response strategies including kill switches for ASI and protocols for rapid containment of a nascent rogue ASI.

Ensure air-gapped networks and robust cyber defence for critical systems to limit a rogue ASI's ability to propagate or gain control.

Prepare resource stockpiles, citizen defence brigades, training programs, and analogue contingency systems to ensure continuity of authority in the event of loss of control to a rogue ASI.