Positive Sum: Technology Alliances for Quantum AI

Amid the concerns and enthusiasm for the memorandum of understanding known as the Technology Prosperity Deal signed in September between the United States and the United Kingdom, , and the simultaneous shift in political relations between the U.S. and European Union as set out in the U.S. National Security Strategy in November, the introduction of a layered approach to tech alliances in quantum AI can strengthen the shared prosperity and security of like-minded nations and build the self-confidence needed to deter adversary influence in an increasingly competitive landscape.  

The pursuit of national interests and collective allied interests are not mutually exclusive. Building strategic, technical alliances in quantum that are targeted and nimble in their aims can serve as a neutral, steadfast vehicle for cooperation and as a form of moral support.  

Research institutions will play a key role in establishing joint initiatives across the artificial intelligence, nuclear energy, and quantum computing fields as announced by the Technology Prosperity Deal. In particular, quantum computing will be advanced through a joint task force of leading researchers to accelerate adoption through industry exchanges in fields like finance, energy, defense, and health care. Moreover, these quantum projects will also explore how to combine the simulation capabilities of quantum computers with the predictive power of AI. 

Securing collaborative wins that showcase the U.S., U.K. and allied leadership in quantum AI (QAI) can contribute to shaping a broader vision for AI – that it is more than just a domineering force for Middle Powers to “catch up with” out of fear from lagging behind. Instead, AI can act as a complementary tool for equally powerful emerging technology, like quantum, in which countries across the EU possess strong leadership positions. Through narrow and specific collaboration, in its most optimistic form, QAI could offer progress for areas of concern, such as user data privacy, through advancements in federated learning in machine learning ecosystems or quantum reinforcement learning that is providing a viable way to decrease electricity costs and power consumption for residential purposes. 

Research: A Key Layer for Strategic Technology Alliances

The research layer of strategic technological alliances represents an emerging strategy for building trust and cooperation. In September 2024, a partnership agreement was formed between military science and technology organizations from the U.S. Defense Advanced Research Projects Agency (DARPA), The U.K. Ministry of Defence’s Science and Technology Laboratory, and the Canadian Department of National Defence’s Research and Development Canada. This agreement demonstrates an original approach with a trilateral international partnership. The effort is designed to both reduce duplication efforts for high-tech research and create interoperable cyber capabilities for cooperation among allies. Synchronizing research developments for AI-based defensive software and network monitoring tools helps to reinforce known vulnerabilities. This is particularly important during a period when both the U.S. and Canada have been targeted by Chinese state-sponsored actors, like the recent Salt Typhoon cyber espionage infiltration across critical telecommunications infrastructure, with a cluster of activity also observed in the U.K.  

While its current efforts focus on AI and cybersecurity, the partnership exemplifies how similar collaborations could be expanded into QAI. Unlike more broad-based security agreements that can become overwhelming in scope or unable to fully obtain a multitude of ambitions, the research layer of strategic technology alliances is narrow, nimble, and neutral in design. 

AI Middle Powers: The Art of Interdependency

Numerous opportunities exist for the U.S. and its allies to work together throughout the AI stack in ways that create a balance and provide key nodes of control in areas of comparative advantage like application development, AI science, research, and startup innovation. Through more strategic political will for technology leadership, the EU can solidify its strengths in critical areas of technology supply chains and academic research as well as reform its weaknesses like the inadequate financial and social bases needed to establish adequate competiveness. For example, by concentrating on adoption and deployment, the U.K. can also seek out open-source strategies to scale AI when appropriate, to prioritize more cost-effective adoption, and reduce unnecessary foreign technology dependencies. 

On the other hand, allies should not remain too focused on how the inability to train frontier AI could symbolize a lack of competitiveness. Instead, they could view the broader picture for how its current internal strengths in areas like quantum technology can ultimately help advance the artificial intelligence landscape in ways that are still receiving little attention. The U.K. ranks second worldwide as a global quantum leader considering the current number of companies and levels of private investment. Similarly, Canada is a world-class base for quantum expertise at the university and industry levels. The EU recognizes the need to put institutional support in place to bring together its AI engineers and quantum physicists in the same room through its Horizon Europe and Quantum Technology Flagship funds. 

The EU ecosystem is expanding with the intent to create more opportunities for this hybrid future. The nascent stage of quantum AI presents an opportunity for international academics and industry to work together. At the pan-European level, the EuroHPC Joint Undertaking provides access to cutting-edge quantum hardware through experimental platforms like the LUMI AI Factory, LUMI-IQ, which combines high-performance computing (HPC) systems, AI, and quantum computing. In addition to platforms, the EuroHPC and six consortia have moved forward with hybrid HPC-QC infrastructure through the designation of six sites across Europe for exploring quantum technologies powered by supercomputers in the EuroQHPC-Integration collaboration.  

Supercomputing, powered by the fastest computers in the world, is synonymous with AI because it offers the high-performance computing needed to run AI workloads. Therefore, hybrid computing systems work together to solve problems “neither can handle alone.” One important role of international research collaboration is to continue to test and evaluate strategies and standards like workflow management systems that can stabilize hardware progress with software and scheduling system evolution. At the research layer, there are numerous opportunities for collaboration with like-minded allies to supercharge innovations and mitigate duplicative efforts. 

QAI: Realities, Expectations, and Challenges

The future between quantum and AI is made up of complementary, and co-existing relationships that can enhance one another in a reciprocal way. The connection between quantum and AI depicts how taking different components from each field and combining them can lead to significant improvements – and that the path toward achieving practical quantum computing may occur through continued experimentation with how the two fields can help one another. Proponents of QAI suggest that with improved quantum hardware, quantum AI models may altogether replace classical systems since they can tackle complexity more efficiently at cheaper cost in terms of energy with scale. This viewpoint also conveys how grand scientific challenges will require integrating quantum technology and artificial intelligence and using these synergistic relationships to their full advantages over time, which is known as quantum accelerated supercomputing.  

Realities

One significant breakthrough from Nvidia’s NVQLink involves combining quantum processors with classical, GPU-based scientific computers in one platform so that both computing resources can be used together to create a more effective outcome. Nvidia’s release of NVQLink provides “an open system architecture” to connect quantum processors and control systems to AI supercomputing in a coherent system. The field of quantum machine learning (QML) is also providing scalable alternatives to traditional methods because quantum features are “used to enhance the learning process,” while the machine learning techniques learn from the “quantum operations to design experiments.” For example, neural networks, a machine learning approach, can be used to detect and identify hidden entanglements created by quantum measurements. These neural networks provide a solution for uncovering connections that previously required scientists to perform time-consuming and repetitive approaches.  

The intention of QML algorithms is to address the disadvantages of using only a single system by leveraging the unique attributes from both quantum systems and machine learning to optimize solutions and provide new ways of processing and representing data. The U.S. Army is already practicing with quantum machine-learning enabled aerial and ground manned and unmanned systems in remote training sites because increased hardware and software capacities are enabling military applications to exponentially enhance speed, security, transmission, and computation. QML is also being used in intelligence, surveillance, and reconnaissance to improve target recognition and anomaly detection. Unique advantages are also being discovered through hybrid classical-quantum approaches like quantum generative models to overcome limitations of classical systems for drug discovery. 

Expectations

Another major advancement of the complementary properties between quantum and AI is the potential for quantum computers to generate new and novel data at a high level of quality for AI to learn from, as AI model training is becoming challenged by a “data wall” from finite supplies of web data. Through hybrid approaches like generative quantum AI (Gen QAI), introduced by Quantinuum, the idea is to “integrate quantum data generation with transformer-based machine learning” to benefit AI system capabilities in areas where training data is limited. Quantum synthetic data can also augment incomplete, real-world datasets needed for mission-critical defense domains that can face collection challenges from intelligence, surveillance, and reconnaissance systems. Quantum synthetic data can provide higher-quality and a higher volume of data than the current capabilities of classical synthetic data generation.  

However, these symbiotic relationships mean that not only can quantum enhance AI, but also AI can enhance quantum. Advances in error correction are also an objective of experimental, early-stage efforts to build more fault-tolerant computation algorithms, which enable a more viable path for practical quantum computing. AI systems can act as a complementary tool to identify errors in quantum computers with high accuracy, which is the objective of Google Quantum AI’s AlphaQubit. In order to predict errors, the tool uses a recurrent, transformer-based, deep learning architecture decoder that learns directly from experimental data in quantum processors. 

Challenges

Despite the proposed 15-year timeline for the emergence of practical quantum computing, there are calls for EU policy to strategically coordinate the support needed to develop AI algorithms, quantum hardware, and enabling technologies. The white paper suggests that to enable this critical co-evolution, more interdisciplinary interaction is needed so research specialists from both quantum and AI can begin to speak the same language and connect with each other, as well as synergize research funding, technology transfer, and talent cultivation. However, for certain types of AI, the application of quantum algorithms faces significant barriers in terms of hardware resources and data-loading issues. Another limitation is that poor, unreliable generalization on quantum devices can occur from the use of AI based methods for preicison sensitive tasks in some instances. 

The Strategic Technology Alliance Trifecta

Three main approaches to strategic technology alliances are emerging: technology development, risk mitigation, and technology transfer. The first two approaches are founded upon a robust research layer between government partners, while the latter is based on a connective layer that enables private sector-led innovation to be transferred to government purposes as needed. The DARPA trilateral partnership for cybersecurity can be categorized as a technological development approach, with the U.K. AI Security Institute (AISI) and U.S. Center for AI Standards and Innovation (CAISI) representing the risk mitigation side. Currently, the AISI and CAISI are not engaged strategically, but part of CAISI’s newly redefined role is to represent U.S. interests internationally.  

Previously, under another name and mission directed toward safety, international leaders agreed to form a network of AI Safety Institutes from the U.K., U.S., and other like-minded allies to share research on risk mitigation like AI model capability evaluations. However, the current White House no longer has any formal or informal intention to continue that mission, although CAISI serves as the primary contact for industry within the U.S. government and will continue to conduct unclassified evaluations of AI capabilities that pose risks to national security, as well as develop and coordinate evaluations and assessments across federal agencies and entities that include security and intelligence communities. 

Safeguarding a networked risk mitigation approach for strategic alliances helps governments better understand the scientific basis of security concerns and the latest tools available to address threats at the frontier. The technological development and risk mitigation approaches are outcomes-based and tend to focus on a narrow and specific area of research, like AI agent-enabled network defense against cyberthreats, as exemplified by the CASTLE program in the DARPA trilateral partnership. 

The Research Layer

Technology joint research at the international level is a new and developing area. By forming a strategic technology alliance based on tech development and risk mitigation, countries can share research efforts in a more secure environment than those that remain at the university level. For example, certain roles with the AISI require national security vetting, and technical positions at DARPA can also require security clearances. In contrast, intelligence has identified university campuses as being “soft targets” for China’s espionage and intellectual property theft.  

A challenge to the technology transfer approach is that the aim of connectivity can be overshadowed by the broader security partnership, so it is important to carve out specific focus areas. One problem experienced with Pillar 2 of the AUKUS security partnership is that it attempted to tackle too many missions, with little joint advancement in areas like the AUKUS Quantum Arrangement. It focused on a wide range of topics, including technological development, research, joint production, and technology transfer AI, quantum, cyber, and undersea. Additionally, sharing tech information on quantum and creating an understanding for regulators at the export control level of dual-use has been extremely difficult because of misinterpretations of how to categorize levels of scruitiny, from scientific advancements to national security concerns. As a result, the AUKUS alliance updated its export controls to remove defense trade barriers for quantum technology development and deployment in defense applications to enable faster and more effective collaboration between partners. 

Overall, AUKUS has evolved into more of an oversight vehicle than one directly coordinating technological advancement. Nonetheless, the technology transfer aspect of Pillar 2 has been successful, especially in industry-led quantum technology. Australia’s Q-CTRL, a quantum software infrastructure firm, has signed contracts with the U.K. Defence Ministry and U.S. Department of Defense Innovation Unit to deliver its quantum sensing technology to prototype a quantum-enabled inertial navigation system. Q-CTRL is also formally registered with the State Department as a participant for eased cross-border transfer purposes. Technology transfer has become easier because of amendments that were made to export controls for quantum between AUKUS countries, such as license-free environments. Other Australian export control change mechanisms are eliminating a significant number of export permits. 

Policy Recommendations

In spite of the fluctuations caused by political disenchantment and the damage inflicted from weaponizing economic choke points, the ability for allies to work together toward neutral, science-led technological progress can build trust, empowerment, and confidence. 

1. Remain open to international cooperation on AI and quantum security. 

Formalizing a research information network between the U.K. AI Security Institute and U.S. CAISI could help set the precedent for future research exchange on risk mitigation for quantum. 

2. Build strategic partnerships for quantum and QAI programs at the university level in allied countries. 

In addition to intergovernmental research lab-based partnerships for emerging technology, other research collaboration structures can be utilized for a more multifaceted approach to strategic research. Choosing a handful of key university players to collaborate on security and defense research projects allows for for more control when awarding multiyear funding by government-based sponsors. An example of how to implement lower-risk university collaboration within an established security partnership can be seen in the Security and Defense PLuS program between Arizona State University, King’s College London, and the University of New South Wales. 

3. Create a research coordination mechanism for QAI. 

Coordinating government research efforts for specific QAI projects may require an intergovernmental committee made up of relevant domestic counterparts to establish plans. This could resemble the Coordinating Committee for Intercontinental Research Networks, which was created in the early days of the internet to coordinate research networks between the EU and North America, eventually expanding to engineering problems. 

The ability to coordinate disparate government and academic based research units across like-minded countries is one way the evolution of the internet took place – from DARPA, to CERN, to the National Science Foundation, to University College London. 

4. Avoid duplicating research efforts across all levels of defense. 

As Europeans take on more leadership within NATO, the United States should refer to already established intelligence-sharing tools, technology, and procedures. The collaborative military capability, Joint Intelligence, Surveillance and Reconnaissance, is an example of how to create a permanent information-sharing system for cooperation on key topics like quantum and QAI where relevant. The United States should keep in mind how financial contributions to NATO defensive efforts can go beyond force posture, such as research initiatives at NATO’s Centre for Maritime Research and Experimentation that developed quantum Mainsail, which uses AI-based machine learning for comprehensive situational awareness in maritime. NATO’s DIANA is providing an accelerator framework where companies are finding solutions to establish transatlantic, quantum-secure communication links between NATO DIANA sites, making progress in quantum safe networks. DIANA is NATO’s analogue of DARPA, so similarities should be used for their cooperative advantages and not overlooked.