Alliances key to AI versus Russia and China---the US can’t do it alone
Andrew Imbrie et al 20, Andrew Imbrie Ryan Fedasiuk Catherine Aiken Tarun Chhabra Husanjot Chahal. Center for Security and Emerging Technology (CSET) at Georgetown’s Walsh School of Foreign Service is a research organization focused on studying the security impacts of emerging technologies, supporting academic work in security and technology studies, and delivering nonpartisan analysis to the policy community. February. "Agile Alliances: How the United States and its Allies Can Deliver a Democratic Way of AI" //pipk
America’s broad network of allies and partners is a source of enduring strength. In an era of economic and technological competition with China and Russia, the United States benefits from allies that share its values and produce troves of strategic resources, including computer and data science experts; private sectors that are innovative, dominant and trend-setting; data on which to train AI algorithms; advanced microprocessors and data storage units; governmental research and development (R&D) investments; diplomatic support for initiatives in AI safety and governance; and the clout needed to export norms and best practices to the rest of the world.
Alliances matter in the AI context because they provide a framework for cooperation, data sharing, dissemination of best practices, joint planning, and procurement. The market adequately incentivizes some forms of cooperation, such as data labelling and exchanges. But alliances can help formalize and expand these relationships, correct for market failures in such areas as AI safety and security, coordinate development of use cases and risk assessments, enhance the legitimacy of international action, and validate the deployment of safe and reliable AI.
America’s alliances and security partnerships will shape the future trajectory of AI, even as AI reshapes the capabilities and operating environments for U.S. allies and partners. By investing in privacy-preserving machine learning and other techniques for improving the interpretability of AI systems, the United States and its allies can promote the development of AI consistent with liberal democratic values. Far-sighted investments could yield large dividends. AI has a wide array of applications that can benefit democracies, from improving data protection and privacy, to promoting transparency and accountability in government.
Advances in AI will also enable new military capabilities. Nations around the world use AI to enhance intelligence collection and analysis, streamline decision-making, lower operating costs, and improve military logistics through predictive maintenance. As China, Russia, and other authoritarian powers integrate AI with military capabilities, U.S. allies and partners will face a more complex operating environment. Advances in software and digital systems could render it more difficult to assess the balance of power in key domains. As the operational tempo of war accelerates, leaders might be tempted to integrate AI and machine learning into early warning and command and control systems, creating new risks and uncertainties for strategic stability.8 Competitors may rush to deploy AI-enabled capabilities without adequate testing, evaluation, verification, and validation. Compounding the risks, adversaries will seek to exploit vulnerabilities in AI systems and may even use AI to execute novel cyberattacks and disinformation campaigns aimed at undermining democratic institutions and sowing discord among the public.9
In meeting these challenges and seizing the opportunities that AI presents, the United States and its allies face tough trade-offs. Three, in particular, necessitate close coordination and prudent mitigation strategies.
First, the United States and its allies face a trade-off between capability and dependency.10 Showcasing a democratic way of AI will require the United States and its allies to pool resources, coordinate policies, and share best practices and information. Leveraging the capabilities of its allies and partners will amplify U.S. power and influence, but will also create inefficiencies and require compromise. While the United States can manage these challenges, it cannot eliminate them entirely—nor should it. As long as AI-related supply chains are global and AI talent both mobile and globally distributed, innovation in AI requires international collaboration.11 To excel in this new context, America will need to embrace its role as a “systems integrator” among like-minded allies and partners.12 Embedding cooperation in dense, decentralized networks plays to the United States’ strengths as a democratic power that favors market approaches to technological cooperation. By combining top-down vision with dynamic, bottom-up innovation and entrepreneurship, the United States and its allies can foster a competitive ecosystem that enables the best ideas to flourish.
These benefits should not obscure the risks. International networks can facilitate cooperation by creating focal points and enhancing the transparency and availability of information.13 As scholars have shown, however, networks of interdependence can also become the sites of competitive power plays, such as the Society for Worldwide Interbank Financial Telecommunication (SWIFT) financial messaging system.14 The United States has used SWIFT to counter terrorism, monitor illicit financial activity, and bolster the sanctions regime against Iran.15 At the same time, China is exploring alternatives to this U.S.-dominated cross-border payments system.16 Emerging competitive dynamics suggest the need for a strategic approach toward the development of AI, especially in semiconductors.
To manage the risks, the United States and its allies should pursue a range of mitigation strategies, including securing and enhancing their defense innovation bases and domestic economic competitiveness, diversifying partnerships in AI, investing diplomatic capacity in norms and standards for AI technologies and mobile telecommunications consistent with democratic values, and promoting flexible institutional configurations for partnership in AI.17
Second, the United States and its allies face a trade-off between competition and cooperation. All nations compete for relative military and economic advantage. In the age of AI, nations will also compete over whose vision of the future attracts the broadest support. Cooperation among democracies is necessary to guard against authoritarian uses of AI, but there are other imperatives for cooperative action. For example, democratic nations must cooperate with strategic competitors to ensure global economic stability and prevent misperceptions and miscalculations from spiraling into hostility; this could be achieved through arms control or international action to create norms and standards for emerging technologies.
Conventional wisdom suggests that the United States competes with China and Russia in AI and collaborates with allies and partners. While accurate, it is equally true that the United States competes with its allies and partners for top talent and resources in AI and must find ways to cooperate with China and Russia on AI safety and security, strategic stability, and crisis management to forestall the risks of accidents and miscalculations.18 When asked about obstacles to collaboration with the United States, for example, multiple officials from allied countries highlighted industrial competition as an impediment.
Cooperative dynamics are typically in pursuit of shared, global interests, while competitive dynamics tend to follow from a national calculation of AI’s impact on relative power and wealth. Researchers and scientists cooperate across national boundaries, but political leaders face difficult trade-offs between national interests and the international networks that foster open-source collaboration in AI.
Navigating these dynamics will require the United States and its allies to pursue a two-pronged strategy: expand areas for cooperation and competition that generate mutual benefits, while shrinking the space for competition that generates harmful effects or a race to the bottom. For example, the United States should manage competition within a rules-based framework that ensures a level playing field, protects intellectual property, and disincentivizes hidden government subsidies. At the same time, it should work with countries to discourage unfair competition that tilts the playing field in favor of state-backed enterprises, destabilizes financial markets, and triggers unforeseen disruptions to global supply chains.
The third trade-off is between safety and speed.19 This trade-off arises from the complex dynamics between the United States and its allies on the one hand, and strategic competitors such as China and Russia on the other. Artificial intelligence presents a range of opportunities and risks for the United States and its allies. AI systems are brittle and can fail accidently or behave unpredictably in real-world settings.20 American, European, Chinese, and Russian leaders increasingly view AI as a core element of national power. In an effort to gain comparative advantage, countries could rush to deploy untested or unsafe AI systems. It is in the interest of U.S. national security to pursue confidence-building measures in AI safety. It is also a core interest of U.S. allies: a majority of officials noted standards to ensure reliable and responsible AI development as a national AI priority and avenue for productive multilateral collaboration. By leading an international effort on safe and reliable AI, the United States and its allies can reduce threats to global security and promote strategic stability.
Policymakers could pursue any number of initiatives in this area, such as bringing together technical experts from the United States, China, and Europe to define shared concepts and standards for the robustness of AI systems; pursuing low-stakes joint projects to summarize the AI safety literature in different countries and promote transparency into applications of AI safety research; facilitating Track 1.5 and Track 2 dialogues on specific challenges in AI safety; or developing common standards and methods of testing, verifying, and validating AI systems, including AI-enabled safety critical infrastructure.21